*eval.txt* For Vim version 9.1. Last change: 2024 Mar 28 VIM REFERENCE MANUAL by Bram Moolenaar Expression evaluation *expression* *expr* *E15* *eval* *E1002* Using expressions is introduced in chapter 41 of the user manual |usr_41.txt|. Note: Expression evaluation can be disabled at compile time. If this has been done, the features in this document are not available. See |+eval| and |no-eval-feature|. This file is mainly about the backwards compatible (legacy) Vim script. For specifics of Vim9 script, which can execute much faster, supports type checking and much more, see |vim9.txt|. Where the syntax or semantics differ a remark is given. 1. Variables |variables| 1.1 Variable types 1.2 Function references |Funcref| 1.3 Lists |Lists| 1.4 Dictionaries |Dictionaries| 1.5 Blobs |Blobs| 1.6 More about variables |more-variables| 2. Expression syntax |expression-syntax| 3. Internal variable |internal-variables| 4. Builtin Functions |functions| 5. Defining functions |user-functions| 6. Curly braces names |curly-braces-names| 7. Commands |expression-commands| 8. Exception handling |exception-handling| 9. Examples |eval-examples| 10. Vim script version |vimscript-version| 11. No +eval feature |no-eval-feature| 12. The sandbox |eval-sandbox| 13. Textlock |textlock| 14. Vim script library |vim-script-library| Testing support is documented in |testing.txt|. Profiling is documented at |profiling|. ============================================================================== 1. Variables *variables* 1.1 Variable types ~ *E712* *E896* *E897* *E899* *E1098* *E1107* *E1135* *E1138* There are ten types of variables: *Number* *Integer* Number A 32 or 64 bit signed number. |expr-number| The number of bits is available in |v:numbersize|. Examples: -123 0x10 0177 0o177 0b1011 Float A floating point number. |floating-point-format| *Float* Examples: 123.456 1.15e-6 -1.1e3 String A NUL terminated string of 8-bit unsigned characters (bytes). |expr-string| Examples: "ab\txx\"--" 'x-z''a,c' List An ordered sequence of items, see |List| for details. Example: [1, 2, ['a', 'b']] Dictionary An associative, unordered array: Each entry has a key and a value. |Dictionary| Examples: {'blue': "#0000ff", 'red': "#ff0000"} #{blue: "#0000ff", red: "#ff0000"} Funcref A reference to a function |Funcref|. Example: function("strlen") It can be bound to a dictionary and arguments, it then works like a Partial. Example: function("Callback", [arg], myDict) Special |v:false|, |v:true|, |v:none| and |v:null|. *Special* Job Used for a job, see |job_start()|. *Job* *Jobs* Channel Used for a channel, see |ch_open()|. *Channel* *Channels* Blob Binary Large Object. Stores any sequence of bytes. See |Blob| for details Example: 0zFF00ED015DAF 0z is an empty Blob. The Number and String types are converted automatically, depending on how they are used. Conversion from a Number to a String is by making the ASCII representation of the Number. Examples: Number 123 --> String "123" ~ Number 0 --> String "0" ~ Number -1 --> String "-1" ~ *octal* Conversion from a String to a Number only happens in legacy Vim script, not in Vim9 script. It is done by converting the first digits to a number. Hexadecimal "0xf9", Octal "017" or "0o17", and Binary "0b10" numbers are recognized NOTE: when using |Vim9| script or |scriptversion-4| octal with a leading "0" is not recognized. The 0o notation requires patch 8.2.0886. If the String doesn't start with digits, the result is zero. Examples: String "456" --> Number 456 ~ String "6bar" --> Number 6 ~ String "foo" --> Number 0 ~ String "0xf1" --> Number 241 ~ String "0100" --> Number 64 ~ String "0o100" --> Number 64 ~ String "0b101" --> Number 5 ~ String "-8" --> Number -8 ~ String "+8" --> Number 0 ~ To force conversion from String to Number, add zero to it: > :echo "0100" + 0 < 64 ~ To avoid a leading zero to cause octal conversion, or for using a different base, use |str2nr()|. *TRUE* *FALSE* *Boolean* For boolean operators Numbers are used. Zero is FALSE, non-zero is TRUE. You can also use |v:false| and |v:true|, in Vim9 script |false| and |true|. When TRUE is returned from a function it is the Number one, FALSE is the number zero. Note that in the command: > :if "foo" :" NOT executed "foo" is converted to 0, which means FALSE. If the string starts with a non-zero number it means TRUE: > :if "8foo" :" executed To test for a non-empty string, use empty(): > :if !empty("foo") < *falsy* *truthy* An expression can be used as a condition, ignoring the type and only using whether the value is "sort of true" or "sort of false". Falsy is: the number zero empty string, blob, list or dictionary Other values are truthy. Examples: 0 falsy 1 truthy -1 truthy 0.0 falsy 0.1 truthy '' falsy 'x' truthy [] falsy [0] truthy {} falsy #{x: 1} truthy 0z falsy 0z00 truthy *non-zero-arg* Function arguments often behave slightly different from |TRUE|: If the argument is present and it evaluates to a non-zero Number, |v:true| or a non-empty String, then the value is considered to be TRUE. Note that " " and "0" are also non-empty strings, thus considered to be TRUE. A List, Dictionary or Float is not a Number or String, thus evaluate to FALSE. *E611* *E745* *E728* *E703* *E729* *E730* *E731* *E908* *E910* *E913* *E974* *E975* *E976* *E1319* *E1320* *E1321* *E1322* *E1323* *E1324* |List|, |Dictionary|, |Funcref|, |Job|, |Channel|, |Blob|, |Class| and |object| types are not automatically converted. *E805* *E806* *E808* When mixing Number and Float the Number is converted to Float. Otherwise there is no automatic conversion of Float. You can use str2float() for String to Float, printf() for Float to String and float2nr() for Float to Number. *E362* *E891* *E892* *E893* *E894* *E907* *E911* *E914* When expecting a Float a Number can also be used, but nothing else. *no-type-checking* You will not get an error if you try to change the type of a variable. 1.2 Function references ~ *Funcref* *E695* *E718* *E1192* A Funcref variable is obtained with the |function()| function, the |funcref()| function, (in |Vim9| script) the name of a function, or created with the lambda expression |expr-lambda|. It can be used in an expression in the place of a function name, before the parenthesis around the arguments, to invoke the function it refers to. Example in |Vim9| script: > :var Fn = MyFunc :echo Fn() Legacy script: > :let Fn = function("MyFunc") :echo Fn() < *E704* *E705* *E707* A Funcref variable must start with a capital, "s:", "w:", "t:" or "b:". You can use "g:" but the following name must still start with a capital. You cannot have both a Funcref variable and a function with the same name. A special case is defining a function and directly assigning its Funcref to a Dictionary entry. Example: > :function dict.init() dict : let self.val = 0 :endfunction The key of the Dictionary can start with a lower case letter. The actual function name is not used here. Also see |numbered-function|. A Funcref can also be used with the |:call| command: > :call Fn() :call dict.init() The name of the referenced function can be obtained with |string()|. > :let func = string(Fn) You can use |call()| to invoke a Funcref and use a list variable for the arguments: > :let r = call(Fn, mylist) < *Partial* A Funcref optionally binds a Dictionary and/or arguments. This is also called a Partial. This is created by passing the Dictionary and/or arguments to function() or funcref(). When calling the function the Dictionary and/or arguments will be passed to the function. Example: > let Cb = function('Callback', ['foo'], myDict) call Cb('bar') This will invoke the function as if using: > call myDict.Callback('foo', 'bar') This is very useful when passing a function around, e.g. in the arguments of |ch_open()|. Note that binding a function to a Dictionary also happens when the function is a member of the Dictionary: > let myDict.myFunction = MyFunction call myDict.myFunction() Here MyFunction() will get myDict passed as "self". This happens when the "myFunction" member is accessed. When making assigning "myFunction" to otherDict and calling it, it will be bound to otherDict: > let otherDict.myFunction = myDict.myFunction call otherDict.myFunction() Now "self" will be "otherDict". But when the dictionary was bound explicitly this won't happen: > let myDict.myFunction = function(MyFunction, myDict) let otherDict.myFunction = myDict.myFunction call otherDict.myFunction() Here "self" will be "myDict", because it was bound explicitly. 1.3 Lists ~ *list* *List* *Lists* *E686* A List is an ordered sequence of items. An item can be of any type. Items can be accessed by their index number. Items can be added and removed at any position in the sequence. List creation ~ *E696* *E697* A List is created with a comma-separated list of items in square brackets. Examples: > :let mylist = [1, two, 3, "four"] :let emptylist = [] An item can be any expression. Using a List for an item creates a List of Lists: > :let nestlist = [[11, 12], [21, 22], [31, 32]] An extra comma after the last item is ignored. List index ~ *list-index* *E684* An item in the List can be accessed by putting the index in square brackets after the List. Indexes are zero-based, thus the first item has index zero. > :let item = mylist[0] " get the first item: 1 :let item = mylist[2] " get the third item: 3 When the resulting item is a list this can be repeated: > :let item = nestlist[0][1] " get the first list, second item: 12 < A negative index is counted from the end. Index -1 refers to the last item in the List, -2 to the last but one item, etc. > :let last = mylist[-1] " get the last item: "four" To avoid an error for an invalid index use the |get()| function. When an item is not available it returns zero or the default value you specify: > :echo get(mylist, idx) :echo get(mylist, idx, "NONE") List concatenation ~ *list-concatenation* Two lists can be concatenated with the "+" operator: > :let longlist = mylist + [5, 6] :let longlist = [5, 6] + mylist To prepend or append an item, turn it into a list by putting [] around it. A list can be concatenated with another one in-place using |:let+=| or |extend()|: > :let mylist += [7, 8] :call extend(mylist, [7, 8]) < See |list-modification| below for more about changing a list in-place. Sublist ~ *sublist* A part of the List can be obtained by specifying the first and last index, separated by a colon in square brackets: > :let shortlist = mylist[2:-1] " get List [3, "four"] Omitting the first index is similar to zero. Omitting the last index is similar to -1. > :let endlist = mylist[2:] " from item 2 to the end: [3, "four"] :let shortlist = mylist[2:2] " List with one item: [3] :let otherlist = mylist[:] " make a copy of the List Notice that the last index is inclusive. If you prefer using an exclusive index use the |slice()| method. If the first index is beyond the last item of the List or the second item is before the first item, the result is an empty list. There is no error message. If the second index is equal to or greater than the length of the list the length minus one is used: > :let mylist = [0, 1, 2, 3] :echo mylist[2:8] " result: [2, 3] NOTE: mylist[s:e] means using the variable "s:e" as index. Watch out for using a single letter variable before the ":". Insert a space when needed: mylist[s : e]. List identity ~ *list-identity* When variable "aa" is a list and you assign it to another variable "bb", both variables refer to the same list. Thus changing the list "aa" will also change "bb": > :let aa = [1, 2, 3] :let bb = aa :call add(aa, 4) :echo bb < [1, 2, 3, 4] Making a copy of a list is done with the |copy()| function. Using [:] also works, as explained above. This creates a shallow copy of the list: Changing a list item in the list will also change the item in the copied list: > :let aa = [[1, 'a'], 2, 3] :let bb = copy(aa) :call add(aa, 4) :let aa[0][1] = 'aaa' :echo aa < [[1, aaa], 2, 3, 4] > :echo bb < [[1, aaa], 2, 3] To make a completely independent list use |deepcopy()|. This also makes a copy of the values in the list, recursively. Up to a hundred levels deep. The operator "is" can be used to check if two variables refer to the same List. "isnot" does the opposite. In contrast "==" compares if two lists have the same value. > :let alist = [1, 2, 3] :let blist = [1, 2, 3] :echo alist is blist < 0 > :echo alist == blist < 1 Note about comparing lists: Two lists are considered equal if they have the same length and all items compare equal, as with using "==". There is one exception: When comparing a number with a string they are considered different. There is no automatic type conversion, as with using "==" on variables. Example: > echo 4 == "4" < 1 > echo [4] == ["4"] < 0 Thus comparing Lists is more strict than comparing numbers and strings. You can compare simple values this way too by putting them in a list: > :let a = 5 :let b = "5" :echo a == b < 1 > :echo [a] == [b] < 0 List unpack ~ To unpack the items in a list to individual variables, put the variables in square brackets, like list items: > :let [var1, var2] = mylist When the number of variables does not match the number of items in the list this produces an error. To handle any extra items from the list append ";" and a variable name: > :let [var1, var2; rest] = mylist This works like: > :let var1 = mylist[0] :let var2 = mylist[1] :let rest = mylist[2:] Except that there is no error if there are only two items. "rest" will be an empty list then. List modification ~ *list-modification* To change a specific item of a list use |:let| this way: > :let list[4] = "four" :let listlist[0][3] = item To change part of a list you can specify the first and last item to be modified. The value must at least have the number of items in the range: > :let list[3:5] = [3, 4, 5] To add items to a List in-place, you can use |:let+=| (|list-concatenation|): > :let listA = [1, 2] :let listA += [3, 4] < When two variables refer to the same List, changing one List in-place will cause the referenced List to be changed in-place: > :let listA = [1, 2] :let listB = listA :let listB += [3, 4] :echo listA [1, 2, 3, 4] < Adding and removing items from a list is done with functions. Here are a few examples: > :call insert(list, 'a') " prepend item 'a' :call insert(list, 'a', 3) " insert item 'a' before list[3] :call add(list, "new") " append String item :call add(list, [1, 2]) " append a List as one new item :call extend(list, [1, 2]) " extend the list with two more items :let i = remove(list, 3) " remove item 3 :unlet list[3] " idem :let l = remove(list, 3, -1) " remove items 3 to last item :unlet list[3 : ] " idem :call filter(list, 'v:val !~ "x"') " remove items with an 'x' Changing the order of items in a list: > :call sort(list) " sort a list alphabetically :call reverse(list) " reverse the order of items :call uniq(sort(list)) " sort and remove duplicates For loop ~ The |:for| loop executes commands for each item in a List, String or Blob. A variable is set to each item in sequence. Example with a List: > :for item in mylist : call Doit(item) :endfor This works like: > :let index = 0 :while index < len(mylist) : let item = mylist[index] : :call Doit(item) : let index = index + 1 :endwhile If all you want to do is modify each item in the list then the |map()| function will be a simpler method than a for loop. Just like the |:let| command, |:for| also accepts a list of variables. This requires the argument to be a List of Lists. > :for [lnum, col] in [[1, 3], [2, 8], [3, 0]] : call Doit(lnum, col) :endfor This works like a |:let| command is done for each list item. Again, the types must remain the same to avoid an error. It is also possible to put remaining items in a List variable: > :for [i, j; rest] in listlist : call Doit(i, j) : if !empty(rest) : echo "remainder: " .. string(rest) : endif :endfor For a Blob one byte at a time is used. For a String one character, including any composing characters, is used as a String. Example: > for c in text echo 'This character is ' .. c endfor List functions ~ *E714* Functions that are useful with a List: > :let r = call(funcname, list) " call a function with an argument list :if empty(list) " check if list is empty :let l = len(list) " number of items in list :let big = max(list) " maximum value in list :let small = min(list) " minimum value in list :let xs = count(list, 'x') " count nr of times 'x' appears in list :let i = index(list, 'x') " index of first 'x' in list :let lines = getline(1, 10) " get ten text lines from buffer :call append('$', lines) " append text lines in buffer :let list = split("a b c") " create list from items in a string :let string = join(list, ', ') " create string from list items :let s = string(list) " String representation of list :call map(list, '">> " .. v:val') " prepend ">> " to each item Don't forget that a combination of features can make things simple. For example, to add up all the numbers in a list: > :exe 'let sum = ' .. join(nrlist, '+') 1.4 Dictionaries ~ *dict* *Dict* *Dictionaries* *Dictionary* A Dictionary is an associative array: Each entry has a key and a value. The entry can be located with the key. The entries are stored without a specific ordering. Dictionary creation ~ *E720* *E721* *E722* *E723* A Dictionary is created with a comma-separated list of entries in curly braces. Each entry has a key and a value, separated by a colon. Each key can only appear once. Examples: > :let mydict = {1: 'one', 2: 'two', 3: 'three'} :let emptydict = {} < *E713* *E716* *E717* A key is always a String. You can use a Number, it will be converted to a String automatically. Thus the String '4' and the number 4 will find the same entry. Note that the String '04' and the Number 04 are different, since the Number will be converted to the String '4', leading zeros are dropped. The empty string can also be used as a key. In |Vim9| script a literal key can be used if it consists only of alphanumeric characters, underscore and dash, see |vim9-literal-dict|. *literal-Dict* *#{}* To avoid having to put quotes around every key the #{} form can be used in legacy script. This does require the key to consist only of ASCII letters, digits, '-' and '_'. Example: > :let mydict = #{zero: 0, one_key: 1, two-key: 2, 333: 3} Note that 333 here is the string "333". Empty keys are not possible with #{}. In |Vim9| script the #{} form cannot be used because it can be confused with the start of a comment. A value can be any expression. Using a Dictionary for a value creates a nested Dictionary: > :let nestdict = {1: {11: 'a', 12: 'b'}, 2: {21: 'c'}} An extra comma after the last entry is ignored. Accessing entries ~ The normal way to access an entry is by putting the key in square brackets: > :let val = mydict["one"] :let mydict["four"] = 4 You can add new entries to an existing Dictionary this way, unlike Lists. For keys that consist entirely of letters, digits and underscore the following form can be used |expr-entry|: > :let val = mydict.one :let mydict.four = 4 Since an entry can be any type, also a List and a Dictionary, the indexing and key lookup can be repeated: > :echo dict.key[idx].key Dictionary to List conversion ~ You may want to loop over the entries in a dictionary. For this you need to turn the Dictionary into a List and pass it to |:for|. Most often you want to loop over the keys, using the |keys()| function: > :for key in keys(mydict) : echo key .. ': ' .. mydict[key] :endfor The List of keys is unsorted. You may want to sort them first: > :for key in sort(keys(mydict)) To loop over the values use the |values()| function: > :for v in values(mydict) : echo "value: " .. v :endfor If you want both the key and the value use the |items()| function. It returns a List in which each item is a List with two items, the key and the value: > :for [key, value] in items(mydict) : echo key .. ': ' .. value :endfor Dictionary identity ~ *dict-identity* Just like Lists you need to use |copy()| and |deepcopy()| to make a copy of a Dictionary. Otherwise, assignment results in referring to the same Dictionary: > :let onedict = {'a': 1, 'b': 2} :let adict = onedict :let adict['a'] = 11 :echo onedict['a'] 11 Two Dictionaries compare equal if all the key-value pairs compare equal. For more info see |list-identity|. Dictionary modification ~ *dict-modification* To change an already existing entry of a Dictionary, or to add a new entry, use |:let| this way: > :let dict[4] = "four" :let dict['one'] = item Removing an entry from a Dictionary is done with |remove()| or |:unlet|. Three ways to remove the entry with key "aaa" from dict: > :let i = remove(dict, 'aaa') :unlet dict.aaa :unlet dict['aaa'] Merging a Dictionary with another is done with |extend()|: > :call extend(adict, bdict) This extends adict with all entries from bdict. Duplicate keys cause entries in adict to be overwritten. An optional third argument can change this. Note that the order of entries in a Dictionary is irrelevant, thus don't expect ":echo adict" to show the items from bdict after the older entries in adict. Weeding out entries from a Dictionary can be done with |filter()|: > :call filter(dict, 'v:val =~ "x"') This removes all entries from "dict" with a value not matching 'x'. This can also be used to remove all entries: > call filter(dict, 0) In some situations it is not allowed to remove or add entries to a Dictionary. Especially when iterating over all the entries. You will get *E1313* or another error in that case. Dictionary function ~ *Dictionary-function* *self* *E725* *E862* When a function is defined with the "dict" attribute it can be used in a special way with a dictionary. Example: > :function Mylen() dict : return len(self.data) :endfunction :let mydict = {'data': [0, 1, 2, 3], 'len': function("Mylen")} :echo mydict.len() This is like a method in object oriented programming. The entry in the Dictionary is a |Funcref|. The local variable "self" refers to the dictionary the function was invoked from. When using |Vim9| script you can use classes and objects, see `:class`. It is also possible to add a function without the "dict" attribute as a Funcref to a Dictionary, but the "self" variable is not available then. *numbered-function* *anonymous-function* To avoid the extra name for the function it can be defined and directly assigned to a Dictionary in this way: > :let mydict = {'data': [0, 1, 2, 3]} :function mydict.len() : return len(self.data) :endfunction :echo mydict.len() The function will then get a number and the value of dict.len is a |Funcref| that references this function. The function can only be used through a |Funcref|. It will automatically be deleted when there is no |Funcref| remaining that refers to it. It is not necessary to use the "dict" attribute for a numbered function. If you get an error for a numbered function, you can find out what it is with a trick. Assuming the function is 42, the command is: > :function g:42 Functions for Dictionaries ~ *E715* Functions that can be used with a Dictionary: > :if has_key(dict, 'foo') " TRUE if dict has entry with key "foo" :if empty(dict) " TRUE if dict is empty :let l = len(dict) " number of items in dict :let big = max(dict) " maximum value in dict :let small = min(dict) " minimum value in dict :let xs = count(dict, 'x') " count nr of times 'x' appears in dict :let s = string(dict) " String representation of dict :call map(dict, '">> " .. v:val') " prepend ">> " to each item 1.5 Blobs ~ *blob* *Blob* *Blobs* *E978* A Blob is a binary object. It can be used to read an image from a file and send it over a channel, for example. A Blob mostly behaves like a |List| of numbers, where each number has the value of an 8-bit byte, from 0 to 255. Blob creation ~ A Blob can be created with a |blob-literal|: > :let b = 0zFF00ED015DAF Dots can be inserted between bytes (pair of hex characters) for readability, they don't change the value: > :let b = 0zFF00.ED01.5DAF A blob can be read from a file with |readfile()| passing the {type} argument set to "B", for example: > :let b = readfile('image.png', 'B') A blob can be read from a channel with the |ch_readblob()| function. Blob index ~ *blob-index* *E979* A byte in the Blob can be accessed by putting the index in square brackets after the Blob. Indexes are zero-based, thus the first byte has index zero. > :let myblob = 0z00112233 :let byte = myblob[0] " get the first byte: 0x00 :let byte = myblob[2] " get the third byte: 0x22 A negative index is counted from the end. Index -1 refers to the last byte in the Blob, -2 to the last but one byte, etc. > :let last = myblob[-1] " get the last byte: 0x33 To avoid an error for an invalid index use the |get()| function. When an item is not available it returns -1 or the default value you specify: > :echo get(myblob, idx) :echo get(myblob, idx, 999) Blob iteration ~ The |:for| loop executes commands for each byte of a Blob. The loop variable is set to each byte in the Blob. Example: > :for byte in 0z112233 : call Doit(byte) :endfor This calls Doit() with 0x11, 0x22 and 0x33. Blob concatenation ~ *blob-concatenation* Two blobs can be concatenated with the "+" operator: > :let longblob = myblob + 0z4455 :let longblob = 0z4455 + myblob < A blob can be concatenated with another one in-place using |:let+=|: > :let myblob += 0z6677 < See |blob-modification| below for more about changing a blob in-place. Part of a blob ~ A part of the Blob can be obtained by specifying the first and last index, separated by a colon in square brackets: > :let myblob = 0z00112233 :let shortblob = myblob[1:2] " get 0z1122 :let shortblob = myblob[2:-1] " get 0z2233 Omitting the first index is similar to zero. Omitting the last index is similar to -1. > :let endblob = myblob[2:] " from item 2 to the end: 0z2233 :let shortblob = myblob[2:2] " Blob with one byte: 0z22 :let otherblob = myblob[:] " make a copy of the Blob If the first index is beyond the last byte of the Blob or the second index is before the first index, the result is an empty Blob. There is no error message. If the second index is equal to or greater than the length of the list the length minus one is used: > :echo myblob[2:8] " result: 0z2233 Blob modification ~ *blob-modification* *E1182* *E1184* To change a specific byte of a blob use |:let| this way: > :let blob[4] = 0x44 When the index is just one beyond the end of the Blob, it is appended. Any higher index is an error. To change a sequence of bytes the [:] notation can be used: > let blob[1:3] = 0z445566 The length of the replaced bytes must be exactly the same as the value provided. *E972* To change part of a blob you can specify the first and last byte to be modified. The value must have the same number of bytes in the range: > :let blob[3:5] = 0z334455 To add items to a Blob in-place, you can use |:let+=| (|blob-concatenation|): > :let blobA = 0z1122 :let blobA += 0z3344 < When two variables refer to the same Blob, changing one Blob in-place will cause the referenced Blob to be changed in-place: > :let blobA = 0z1122 :let blobB = blobA :let blobB += 0z3344 :echo blobA 0z11223344 < You can also use the functions |add()|, |remove()| and |insert()|. Blob identity ~ Blobs can be compared for equality: > if blob == 0z001122 And for equal identity: > if blob is otherblob < *blob-identity* *E977* When variable "aa" is a Blob and you assign it to another variable "bb", both variables refer to the same Blob. Then the "is" operator returns true. When making a copy using [:] or |copy()| the values are the same, but the identity is different: > :let blob = 0z112233 :let blob2 = blob :echo blob == blob2 < 1 > :echo blob is blob2 < 1 > :let blob3 = blob[:] :echo blob == blob3 < 1 > :echo blob is blob3 < 0 Making a copy of a Blob is done with the |copy()| function. Using [:] also works, as explained above. 1.6 More about variables ~ *more-variables* If you need to know the type of a variable or expression, use the |type()| function. When the '!' flag is included in the 'viminfo' option, global variables that start with an uppercase letter, and don't contain a lowercase letter, are stored in the viminfo file |viminfo-file|. When the 'sessionoptions' option contains "global", global variables that start with an uppercase letter and contain at least one lowercase letter are stored in the session file |session-file|. variable name can be stored where ~ my_var_6 not My_Var_6 session file MY_VAR_6 viminfo file In legacy script it is possible to form a variable name with curly braces, see |curly-braces-names|. ============================================================================== 2. Expression syntax *expression-syntax* *E1143* Expression syntax summary, from least to most significant: |expr1| expr2 expr2 ? expr1 : expr1 if-then-else |expr2| expr3 expr3 || expr3 ... logical OR |expr3| expr4 expr4 && expr4 ... logical AND |expr4| expr5 expr5 == expr5 equal expr5 != expr5 not equal expr5 > expr5 greater than expr5 >= expr5 greater than or equal expr5 < expr5 smaller than expr5 <= expr5 smaller than or equal expr5 =~ expr5 regexp matches expr5 !~ expr5 regexp doesn't match expr5 ==? expr5 equal, ignoring case expr5 ==# expr5 equal, match case etc. As above, append ? for ignoring case, # for matching case expr5 is expr5 same |List|, |Dictionary| or |Blob| instance expr5 isnot expr5 different |List|, |Dictionary| or |Blob| instance |expr5| expr6 expr6 << expr6 bitwise left shift expr6 >> expr6 bitwise right shift |expr6| expr7 expr7 + expr7 ... number addition, list or blob concatenation expr7 - expr7 ... number subtraction expr7 . expr7 ... string concatenation expr7 .. expr7 ... string concatenation |expr7| expr8 expr8 * expr8 ... number multiplication expr8 / expr8 ... number division expr8 % expr8 ... number modulo |expr8| expr9 expr9 type check and conversion (|Vim9| only) |expr9| expr10 ! expr9 logical NOT - expr9 unary minus + expr9 unary plus |expr10| expr11 expr10[expr1] byte of a String or item of a |List| expr10[expr1 : expr1] substring of a String or sublist of a |List| expr10.name entry in a |Dictionary| expr10(expr1, ...) function call with |Funcref| variable expr10->name(expr1, ...) |method| call |expr11| number number constant "string" string constant, backslash is special 'string' string constant, ' is doubled [expr1, ...] |List| {expr1: expr1, ...} |Dictionary| #{key: expr1, ...} legacy |Dictionary| &option option value (expr1) nested expression variable internal variable va{ria}ble internal variable with curly braces $VAR environment variable @r contents of register 'r' function(expr1, ...) function call func{ti}on(expr1, ...) function call with curly braces {args -> expr1} legacy lambda expression (args) => expr1 Vim9 lambda expression "..." indicates that the operations in this level can be concatenated. Example: > &nu || &list && &shell == "csh" All expressions within one level are parsed from left to right. Expression nesting is limited to 1000 levels deep (300 when build with MSVC) to avoid running out of stack and crashing. *E1169* expr1 *expr1* *ternary* *falsy-operator* *??* *E109* ----- The ternary operator: expr2 ? expr1 : expr1 The falsy operator: expr2 ?? expr1 Ternary operator ~ In legacy script the expression before the '?' is evaluated to a number. If it evaluates to |TRUE|, the result is the value of the expression between the '?' and ':', otherwise the result is the value of the expression after the ':'. In |Vim9| script the first expression must evaluate to a boolean, see |vim9-boolean|. Example: > :echo lnum == 1 ? "top" : lnum Since the first expression is an "expr2", it cannot contain another ?:. The other two expressions can, thus allow for recursive use of ?:. Example: > :echo lnum == 1 ? "top" : lnum == 1000 ? "last" : lnum To keep this readable, using |line-continuation| is suggested: > :echo lnum == 1 :\ ? "top" :\ : lnum == 1000 :\ ? "last" :\ : lnum You should always put a space before the ':', otherwise it can be mistaken for use in a variable such as "a:1". Falsy operator ~ This is also known as the "null coalescing operator", but that's too complicated, thus we just call it the falsy operator. The expression before the '??' is evaluated. If it evaluates to |truthy|, this is used as the result. Otherwise the expression after the '??' is evaluated and used as the result. This is most useful to have a default value for an expression that may result in zero or empty: > echo theList ?? 'list is empty' echo GetName() ?? 'unknown' These are similar, but not equal: > expr2 ?? expr1 expr2 ? expr2 : expr1 In the second line "expr2" is evaluated twice. And in |Vim9| script the type of expr2 before "?" must be a boolean. expr2 and expr3 *expr2* *expr3* --------------- expr3 || expr3 .. logical OR *expr-barbar* expr4 && expr4 .. logical AND *expr-&&* The "||" and "&&" operators take one argument on each side. In legacy script the arguments are (converted to) Numbers. In |Vim9| script the values must be boolean, see |vim9-boolean|. Use "!!" to convert any type to a boolean. The result is: input output ~ n1 n2 n1 || n2 n1 && n2 ~ |FALSE| |FALSE| |FALSE| |FALSE| |FALSE| |TRUE| |TRUE| |FALSE| |TRUE| |FALSE| |TRUE| |FALSE| |TRUE| |TRUE| |TRUE| |TRUE| The operators can be concatenated, for example: > &nu || &list && &shell == "csh" Note that "&&" takes precedence over "||", so this has the meaning of: > &nu || (&list && &shell == "csh") Once the result is known, the expression "short-circuits", that is, further arguments are not evaluated. This is like what happens in C. For example: > let a = 1 echo a || b This is valid even if there is no variable called "b" because "a" is |TRUE|, so the result must be |TRUE|. Similarly below: > echo exists("b") && b == "yes" This is valid whether "b" has been defined or not. The second clause will only be evaluated if "b" has been defined. expr4 *expr4* *E1153* ----- expr5 {cmp} expr5 Compare two expr5 expressions. In legacy script the result is a 0 if it evaluates to false, or 1 if it evaluates to true. In |Vim9| script the result is |true| or |false|. *expr-==* *expr-!=* *expr->* *expr->=* *expr-<* *expr-<=* *expr-=~* *expr-!~* *expr-==#* *expr-!=#* *expr->#* *expr->=#* *expr-<#* *expr-<=#* *expr-=~#* *expr-!~#* *expr-==?* *expr-!=?* *expr->?* *expr->=?* *expr- ># >? greater than or equal >= >=# >=? smaller than < <# if get(Part1, 'name') == get(Part2, 'name') " Part1 and Part2 refer to the same function < *E1037* Using "is" or "isnot" with a |List|, |Dictionary| or |Blob| checks whether the expressions are referring to the same |List|, |Dictionary| or |Blob| instance. A copy of a |List| is different from the original |List|. When using "is" without a |List|, |Dictionary| or |Blob|, it is equivalent to using "equal", using "isnot" equivalent to using "not equal". Except that a different type means the values are different: > echo 4 == '4' 1 echo 4 is '4' 0 echo 0 is [] 0 "is#"/"isnot#" and "is?"/"isnot?" can be used to match and ignore case. In |Vim9| script this doesn't work, two strings are never identical. In legacy script, when comparing a String with a Number, the String is converted to a Number, and the comparison is done on Numbers. This means that: > echo 0 == 'x' 1 because 'x' converted to a Number is zero. However: > echo [0] == ['x'] 0 Inside a List or Dictionary this conversion is not used. In |Vim9| script the types must match. When comparing two Strings, this is done with strcmp() or stricmp(). This results in the mathematical difference (comparing byte values), not necessarily the alphabetical difference in the local language. When using the operators with a trailing '#', or the short version and 'ignorecase' is off, the comparing is done with strcmp(): case matters. When using the operators with a trailing '?', or the short version and 'ignorecase' is set, the comparing is done with stricmp(): case is ignored. 'smartcase' is not used. The "=~" and "!~" operators match the lefthand argument with the righthand argument, which is used as a pattern. See |pattern| for what a pattern is. This matching is always done like 'magic' was set and 'cpoptions' is empty, no matter what the actual value of 'magic' or 'cpoptions' is. This makes scripts portable. To avoid backslashes in the regexp pattern to be doubled, use a single-quote string, see |literal-string|. Since a string is considered to be a single line, a multi-line pattern (containing \n, backslash-n) will not match. However, a literal NL character can be matched like an ordinary character. Examples: "foo\nbar" =~ "\n" evaluates to 1 "foo\nbar" =~ "\\n" evaluates to 0 expr5 *expr5* *bitwise-shift* ----- expr6 << expr6 bitwise left shift *expr-<<* expr6 >> expr6 bitwise right shift *expr->>* *E1282* *E1283* The "<<" and ">>" operators can be used to perform bitwise left or right shift of the left operand by the number of bits specified by the right operand. The operands are used as positive numbers. When shifting right with ">>" the topmost bit (sometimes called the sign bit) is cleared. If the right operand (shift amount) is more than the maximum number of bits in a number (|v:numbersize|) the result is zero. expr6 and expr7 *expr6* *expr7* *E1036* *E1051* --------------- expr7 + expr7 Number addition, |List| or |Blob| concatenation *expr-+* expr7 - expr7 Number subtraction *expr--* expr7 . expr7 String concatenation *expr-.* expr7 .. expr7 String concatenation *expr-..* For |Lists| only "+" is possible and then both expr7 must be a list. The result is a new list with the two lists Concatenated. For String concatenation ".." is preferred, since "." is ambiguous, it is also used for |Dict| member access and floating point numbers. In |Vim9| script and when |vimscript-version| is 2 or higher, using "." is not allowed. In |Vim9| script the arguments of ".." are converted to String for simple types: Number, Float, Special and Bool. For other types |string()| should be used. expr8 * expr8 Number multiplication *expr-star* expr8 / expr8 Number division *expr-/* expr8 % expr8 Number modulo *expr-%* In legacy script, for all operators except "." and "..", Strings are converted to Numbers. For bitwise operators see |and()|, |or()| and |xor()|. Note the difference between "+" and ".." in legacy script: "123" + "456" = 579 "123" .. "456" = "123456" Since '..' has the same precedence as '+' and '-', you need to read: > 1 .. 90 + 90.0 As: > (1 .. 90) + 90.0 That works in legacy script, since the String "190" is automatically converted to the Number 190, which can be added to the Float 90.0. However: > 1 .. 90 * 90.0 Should be read as: > 1 .. (90 * 90.0) Since '..' has lower precedence than '*'. This does NOT work, since this attempts to concatenate a Float and a String. When dividing a Number by zero the result depends on the value: 0 / 0 = -0x80000000 (like NaN for Float) >0 / 0 = 0x7fffffff (like positive infinity) <0 / 0 = -0x7fffffff (like negative infinity) (before Vim 7.2 it was always 0x7fffffff) In |Vim9| script dividing a number by zero is an error. *E1154* When 64-bit Number support is enabled: 0 / 0 = -0x8000000000000000 (like NaN for Float) >0 / 0 = 0x7fffffffffffffff (like positive infinity) <0 / 0 = -0x7fffffffffffffff (like negative infinity) When the righthand side of '%' is zero, the result is 0. None of these work for |Funcref|s. ".", ".." and "%" do not work for Float. *E804* *E1035* expr8 *expr8* ----- expr9 This is only available in |Vim9| script, see |type-casting|. expr9 *expr9* ----- ! expr9 logical NOT *expr-!* - expr9 unary minus *expr-unary--* + expr9 unary plus *expr-unary-+* For '!' |TRUE| becomes |FALSE|, |FALSE| becomes |TRUE| (one). For '-' the sign of the number is changed. For '+' the number is unchanged. Note: "++" has no effect. In legacy script a String will be converted to a Number first. Note that if the string does not start with a digit you likely don't get what you expect. In |Vim9| script an error is given when "-" or "+" is used and the type is not a number. In |Vim9| script "!" can be used for any type and the result is always a boolean. Use "!!" to convert any type to a boolean, according to whether the value is |falsy|. These three can be repeated and mixed. Examples: !-1 == 0 !!8 == 1 --9 == 9 expr10 *expr10* ------ This expression is either |expr11| or a sequence of the alternatives below, in any order. E.g., these are all possible: expr10[expr1].name expr10.name[expr1] expr10(expr1, ...)[expr1].name expr10->(expr1, ...)[expr1] Evaluation is always from left to right. expr10[expr1] item of String or |List| *expr-[]* *E111* *E909* *subscript* *E1062* In legacy Vim script: If expr10 is a Number or String this results in a String that contains the expr1'th single byte from expr10. expr10 is used as a String (a number is automatically converted to a String), expr1 as a Number. This doesn't recognize multibyte encodings, see `byteidx()` for an alternative, or use `split()` to turn the string into a list of characters. Example, to get the byte under the cursor: > :let c = getline(".")[col(".") - 1] In |Vim9| script: *E1147* *E1148* If expr10 is a String this results in a String that contains the expr1'th single character (including any composing characters) from expr10. To use byte indexes use |strpart()|. Index zero gives the first byte or character. Careful: text column numbers start with one! If the length of the String is less than the index, the result is an empty String. A negative index always results in an empty string (reason: backward compatibility). Use [-1:] to get the last byte or character. In Vim9 script a negative index is used like with a list: count from the end. If expr10 is a |List| then it results the item at index expr1. See |list-index| for possible index values. If the index is out of range this results in an error. Example: > :let item = mylist[-1] " get last item Generally, if a |List| index is equal to or higher than the length of the |List|, or more negative than the length of the |List|, this results in an error. expr10[expr1a : expr1b] substring or |sublist| *expr-[:]* *substring* If expr10 is a String this results in the substring with the bytes or characters from expr1a to and including expr1b. expr10 is used as a String, expr1a and expr1b are used as a Number. In legacy Vim script the indexes are byte indexes. This doesn't recognize multibyte encodings, see |byteidx()| for computing the indexes. If expr10 is a Number it is first converted to a String. In Vim9 script the indexes are character indexes and include composing characters. To use byte indexes use |strpart()|. To use character indexes without including composing characters use |strcharpart()|. The item at index expr1b is included, it is inclusive. For an exclusive index use the |slice()| function. If expr1a is omitted zero is used. If expr1b is omitted the length of the string minus one is used. A negative number can be used to measure from the end of the string. -1 is the last character, -2 the last but one, etc. If an index goes out of range for the string characters are omitted. If expr1b is smaller than expr1a the result is an empty string. Examples: > :let c = name[-1:] " last byte of a string :let c = name[0:-1] " the whole string :let c = name[-2:-2] " last but one byte of a string :let s = line(".")[4:] " from the fifth byte to the end :let s = s[:-3] " remove last two bytes < *slice* If expr10 is a |List| this results in a new |List| with the items indicated by the indexes expr1a and expr1b. This works like with a String, as explained just above. Also see |sublist| below. Examples: > :let l = mylist[:3] " first four items :let l = mylist[4:4] " List with one item :let l = mylist[:] " shallow copy of a List If expr10 is a |Blob| this results in a new |Blob| with the bytes in the indexes expr1a and expr1b, inclusive. Examples: > :let b = 0zDEADBEEF :let bs = b[1:2] " 0zADBE :let bs = b[:] " copy of 0zDEADBEEF Using expr10[expr1] or expr10[expr1a : expr1b] on a |Funcref| results in an error. Watch out for confusion between a namespace and a variable followed by a colon for a sublist: > mylist[n:] " uses variable n mylist[s:] " uses namespace s:, error! expr10.name entry in a |Dictionary| *expr-entry* *E1203* *E1229* If expr10 is a |Dictionary| and it is followed by a dot, then the following name will be used as a key in the |Dictionary|. This is just like: expr10[name]. The name must consist of alphanumeric characters, just like a variable name, but it may start with a number. Curly braces cannot be used. There must not be white space before or after the dot. Examples: > :let dict = {"one": 1, 2: "two"} :echo dict.one " shows "1" :echo dict.2 " shows "two" :echo dict .2 " error because of space before the dot Note that the dot is also used for String concatenation. To avoid confusion always put spaces around the dot for String concatenation. expr10(expr1, ...) |Funcref| function call *E1085* When expr10 is a |Funcref| type variable, invoke the function it refers to. expr10->name([args]) method call *method* *->* expr10->{lambda}([args]) *E260* *E276* *E1265* For methods that are also available as global functions this is the same as: > name(expr10 [, args]) There can also be methods specifically for the type of "expr10". This allows for chaining, passing the value that one method returns to the next method: > mylist->filter(filterexpr)->map(mapexpr)->sort()->join() < Example of using a lambda: > GetPercentage()->{x -> x * 100}()->printf('%d%%') < When using -> the |expr9| operators will be applied first, thus: > -1.234->string() Is equivalent to: > (-1.234)->string() And NOT: > -(1.234->string()) What comes after "->" can be a name, a simple expression (not containing any parenthesis), or any expression in parentheses: > base->name(args) base->some.name(args) base->alist[idx](args) base->(getFuncRef())(args) Note that in the last call the base is passed to the function resulting from "(getFuncRef())", inserted before "args". *E1275* *E274* "->name(" must not contain white space. There can be white space before the "->" and after the "(", thus you can split the lines like this: > mylist \ ->filter(filterexpr) \ ->map(mapexpr) \ ->sort() \ ->join() When using the lambda form there must be no white space between the } and the (. *expr11* number ------ number number constant *expr-number* *0x* *hex-number* *0o* *octal-number* *binary-number* Decimal, Hexadecimal (starting with 0x or 0X), Binary (starting with 0b or 0B) and Octal (starting with 0, 0o or 0O). Assuming 64 bit numbers are used (see |v:numbersize|) an unsigned number is truncated to 0x7fffffffffffffff or 9223372036854775807. You can use -1 to get 0xffffffffffffffff. *floating-point-format* Floating point numbers can be written in two forms: [-+]{N}.{M} [-+]{N}.{M}[eE][-+]{exp} {N} and {M} are numbers. Both {N} and {M} must be present and can only contain digits, except that in |Vim9| script in {N} single quotes between digits are ignored. [-+] means there is an optional plus or minus sign. {exp} is the exponent, power of 10. Only a decimal point is accepted, not a comma. No matter what the current locale is. Examples: 123.456 +0.0001 55.0 -0.123 1.234e03 1.0E-6 -3.1416e+88 These are INVALID: 3. empty {M} 1e40 missing .{M} Rationale: Before floating point was introduced, the text "123.456" was interpreted as the two numbers "123" and "456", both converted to a string and concatenated, resulting in the string "123456". Since this was considered pointless, and we could not find it intentionally being used in Vim scripts, this backwards incompatibility was accepted in favor of being able to use the normal notation for floating point numbers. *float-pi* *float-e* A few useful values to copy&paste: > :let pi = 3.14159265359 :let e = 2.71828182846 Or, if you don't want to write them in as floating-point literals, you can also use functions, like the following: > :let pi = acos(-1.0) :let e = exp(1.0) < *floating-point-precision* The precision and range of floating points numbers depends on what "double" means in the library Vim was compiled with. There is no way to change this at runtime. The default for displaying a |Float| is to use 6 decimal places, like using printf("%g", f). You can select something else when using the |printf()| function. Example: > :echo printf('%.15e', atan(1)) < 7.853981633974483e-01 string *string* *String* *expr-string* *E114* ------ "string" string constant *expr-quote* Note that double quotes are used. A string constant accepts these special characters: \... three-digit octal number (e.g., "\316") \.. two-digit octal number (must be followed by non-digit) \. one-digit octal number (must be followed by non-digit) \x.. byte specified with two hex numbers (e.g., "\x1f") \x. byte specified with one hex number (must be followed by non-hex char) \X.. same as \x.. \X. same as \x. \u.... character specified with up to 4 hex numbers, stored according to the current value of 'encoding' (e.g., "\u02a4") \U.... same as \u but allows up to 8 hex numbers. \b backspace \e escape \f formfeed 0x0C \n newline \r return \t tab \\ backslash \" double quote \ Special key named "xxx". e.g. "\" for CTRL-W. This is for use in mappings, the 0x80 byte is escaped. To use the double quote character it must be escaped: "". Don't use to get a UTF-8 character, use \uxxxx as mentioned above. \<*xxx> Like \ but prepends a modifier instead of including it in the character. E.g. "\" is one character 0x17 while "\<*C-w>" is four bytes: 3 for the CTRL modifier and then character "W". Note that "\xff" is stored as the byte 255, which may be invalid in some encodings. Use "\u00ff" to store character 255 according to the current value of 'encoding'. Note that "\000" and "\x00" force the end of the string. blob-literal *blob-literal* *E973* ------------ Hexadecimal starting with 0z or 0Z, with an arbitrary number of bytes. The sequence must be an even number of hex characters. Example: > :let b = 0zFF00ED015DAF literal-string *literal-string* *E115* --------------- 'string' string constant *expr-'* Note that single quotes are used. This string is taken as it is. No backslashes are removed or have a special meaning. The only exception is that two quotes stand for one quote. Single quoted strings are useful for patterns, so that backslashes do not need to be doubled. These two commands are equivalent: > if a =~ "\\s*" if a =~ '\s*' interpolated-string *$quote* *interpolated-string* -------------------- $"string" interpolated string constant *expr-$quote* $'string' interpolated literal string constant *expr-$'* Interpolated strings are an extension of the |string| and |literal-string|, allowing the inclusion of Vim script expressions (see |expr1|). Any expression returning a value can be enclosed between curly braces. The value is converted to a string. All the text and results of the expressions are concatenated to make a new string. *E1278* *E1279* To include an opening brace '{' or closing brace '}' in the string content double it. For double quoted strings using a backslash also works. A single closing brace '}' will result in an error. Examples: > let your_name = input("What's your name? ") < What's your name? Peter ~ > echo echo $"Hello, {your_name}!" < Hello, Peter! ~ > echo $"The square root of {{9}} is {sqrt(9)}" < The square root of {9} is 3.0 ~ *string-offset-encoding* A string consists of multiple characters. How the characters are stored depends on 'encoding'. Most common is UTF-8, which uses one byte for ASCII characters, two bytes for other latin characters and more bytes for other characters. A string offset can count characters or bytes. Other programs may use UTF-16 encoding (16-bit words) and an offset of UTF-16 words. Some functions use byte offsets, usually for UTF-8 encoding. Other functions use character offsets, in which case the encoding doesn't matter. The different offsets for the string "a©😊" are below: UTF-8 offsets: [0]: 61, [1]: C2, [2]: A9, [3]: F0, [4]: 9F, [5]: 98, [6]: 8A UTF-16 offsets: [0]: 0061, [1]: 00A9, [2]: D83D, [3]: DE0A UTF-32 (character) offsets: [0]: 00000061, [1]: 000000A9, [2]: 0001F60A You can use the "g8" and "ga" commands on a character to see the decimal/hex/octal values. The functions |byteidx()|, |utf16idx()| and |charidx()| can be used to convert between these indices. The functions |strlen()|, |strutf16len()| and |strcharlen()| return the number of bytes, UTF-16 code units and characters in a string respectively. option *expr-option* *E112* *E113* ------ &option option value, local value if possible &g:option global option value &l:option local option value Examples: > echo "tabstop is " .. &tabstop if &insertmode Any option name can be used here. See |options|. When using the local value and there is no buffer-local or window-local value, the global value is used anyway. register *expr-register* *@r* -------- @r contents of register 'r' The result is the contents of the named register, as a single string. Newlines are inserted where required. To get the contents of the unnamed register use @" or @@. See |registers| for an explanation of the available registers. When using the '=' register you get the expression itself, not what it evaluates to. Use |eval()| to evaluate it. nesting *expr-nesting* *E110* ------- (expr1) nested expression environment variable *expr-env* -------------------- $VAR environment variable The String value of any environment variable. When it is not defined, the result is an empty string. The functions `getenv()` and `setenv()` can also be used and work for environment variables with non-alphanumeric names. The function `environ()` can be used to get a Dict with all environment variables. *expr-env-expand* Note that there is a difference between using $VAR directly and using expand("$VAR"). Using it directly will only expand environment variables that are known inside the current Vim session. Using expand() will first try using the environment variables known inside the current Vim session. If that fails, a shell will be used to expand the variable. This can be slow, but it does expand all variables that the shell knows about. Example: > :echo $shell :echo expand("$shell") The first one probably doesn't echo anything, the second echoes the $shell variable (if your shell supports it). internal variable *expr-variable* *E1015* *E1089* ----------------- variable internal variable See below |internal-variables|. function call *expr-function* *E116* *E118* *E119* *E120* ------------- function(expr1, ...) function call See below |functions|. lambda expression *expr-lambda* *lambda* ----------------- {args -> expr1} legacy lambda expression *E451* (args) => expr1 |Vim9| lambda expression A lambda expression creates a new unnamed function which returns the result of evaluating |expr1|. Lambda expressions differ from |user-functions| in the following ways: 1. The body of the lambda expression is an |expr1| and not a sequence of |Ex| commands. 2. The prefix "a:" should not be used for arguments. E.g.: > :let F = {arg1, arg2 -> arg1 - arg2} :echo F(5, 2) < 3 The arguments are optional. Example: > :let F = {-> 'error function'} :echo F('ignored') < error function The |Vim9| lambda does not only use a different syntax, it also adds type checking and can be split over multiple lines, see |vim9-lambda|. *closure* Lambda expressions can access outer scope variables and arguments. This is often called a closure. Example where "i" and "a:arg" are used in a lambda while they already exist in the function scope. They remain valid even after the function returns: > :function Foo(arg) : let i = 3 : return {x -> x + i - a:arg} :endfunction :let Bar = Foo(4) :echo Bar(6) < 5 Note that the variables must exist in the outer scope before the lambda is defined for this to work. See also |:func-closure|. Lambda and closure support can be checked with: > if has('lambda') Examples for using a lambda expression with |sort()|, |map()| and |filter()|: > :echo map([1, 2, 3], {idx, val -> val + 1}) < [2, 3, 4] > :echo sort([3,7,2,1,4], {a, b -> a - b}) < [1, 2, 3, 4, 7] The lambda expression is also useful for Channel, Job and timer: > :let timer = timer_start(500, \ {-> execute("echo 'Handler called'", "")}, \ {'repeat': 3}) < Handler called Handler called Handler called Note that it is possible to cause memory to be used and not freed if the closure is referenced by the context it depends on: > function Function() let x = 0 let F = {-> x} endfunction The closure uses "x" from the function scope, and "F" in that same scope refers to the closure. This cycle results in the memory not being freed. Recommendation: don't do this. Notice how execute() is used to execute an Ex command. That's ugly though. In Vim9 script you can use a command block, see |inline-function|. Although you can use the loop variable of a `for` command, it must still exist when the closure is called, otherwise you get an error. *E1302* Lambda expressions have internal names like '42'. If you get an error for a lambda expression, you can find what it is with the following command: > :function 42 See also: |numbered-function| ============================================================================== 3. Internal variable *internal-variables* *E461* *E1001* An internal variable name can be made up of letters, digits and '_'. But it cannot start with a digit. In legacy script it is also possible to use curly braces, see |curly-braces-names|. In legacy script an internal variable is created with the ":let" command |:let|. An internal variable is explicitly destroyed with the ":unlet" command |:unlet|. Using a name that is not an internal variable or refers to a variable that has been destroyed results in an error. In |Vim9| script `:let` is not used and variables work differently, see |:var|. *variable-scope* There are several name spaces for variables. Which one is to be used is specified by what is prepended: (nothing) In a function: local to the function; in a legacy script: global; in a |Vim9| script: local to the script |buffer-variable| b: Local to the current buffer. |window-variable| w: Local to the current window. |tabpage-variable| t: Local to the current tab page. |global-variable| g: Global. |local-variable| l: Local to a function (only in a legacy function) |script-variable| s: Local to a |:source|'ed Vim script. |function-argument| a: Function argument (only in a legacy function). |vim-variable| v: Global, predefined by Vim. The scope name by itself can be used as a |Dictionary|. For example, to delete all script-local variables: > :for k in keys(s:) : unlet s:[k] :endfor Note: in Vim9 script variables can also be local to a block of commands, see |vim9-scopes|. *buffer-variable* *b:var* *b:* A variable name that is preceded with "b:" is local to the current buffer. Thus you can have several "b:foo" variables, one for each buffer. This kind of variable is deleted when the buffer is wiped out or deleted with |:bdelete|. One local buffer variable is predefined: *b:changedtick* *changetick* b:changedtick The total number of changes to the current buffer. It is incremented for each change. An undo command is also a change in this case. Resetting 'modified' when writing the buffer is also counted. This can be used to perform an action only when the buffer has changed. Example: > :if my_changedtick != b:changedtick : let my_changedtick = b:changedtick : call My_Update() :endif < You cannot change or delete the b:changedtick variable. If you need more information about the change see |listener_add()|. *window-variable* *w:var* *w:* A variable name that is preceded with "w:" is local to the current window. It is deleted when the window is closed. *tabpage-variable* *t:var* *t:* A variable name that is preceded with "t:" is local to the current tab page, It is deleted when the tab page is closed. {not available when compiled without the |+windows| feature} *global-variable* *g:var* *g:* Inside functions and in |Vim9| script global variables are accessed with "g:". Omitting this will access a variable local to a function or script. "g:" can also be used in any other place if you like. *local-variable* *l:var* *l:* Inside functions local variables are accessed without prepending anything. But you can also prepend "l:" if you like. However, without prepending "l:" you may run into reserved variable names. For example "count". By itself it refers to "v:count". Using "l:count" you can have a local variable with the same name. *script-variable* *s:var* In a legacy Vim script variables starting with "s:" can be used. They cannot be accessed from outside of the scripts, thus are local to the script. In |Vim9| script the "s:" prefix can be omitted, variables are script-local by default. They can be used in: - commands executed while the script is sourced - functions defined in the script - autocommands defined in the script - functions and autocommands defined in functions and autocommands which were defined in the script (recursively) - user defined commands defined in the script Thus not in: - other scripts sourced from this one - mappings - menus - etc. Script variables can be used to avoid conflicts with global variable names. Take this example: > let s:counter = 0 function MyCounter() let s:counter = s:counter + 1 echo s:counter endfunction command Tick call MyCounter() You can now invoke "Tick" from any script, and the "s:counter" variable in that script will not be changed, only the "s:counter" in the script where "Tick" was defined is used. Another example that does the same: > let s:counter = 0 command Tick let s:counter = s:counter + 1 | echo s:counter When calling a function and invoking a user-defined command, the context for script variables is set to the script where the function or command was defined. The script variables are also available when a function is defined inside a function that is defined in a script. Example: > let s:counter = 0 function StartCounting(incr) if a:incr function MyCounter() let s:counter = s:counter + 1 endfunction else function MyCounter() let s:counter = s:counter - 1 endfunction endif endfunction This defines the MyCounter() function either for counting up or counting down when calling StartCounting(). It doesn't matter from where StartCounting() is called, the s:counter variable will be accessible in MyCounter(). When the same script is sourced again it will use the same script variables. They will remain valid as long as Vim is running. This can be used to maintain a counter: > if !exists("s:counter") let s:counter = 1 echo "script executed for the first time" else let s:counter = s:counter + 1 echo "script executed " .. s:counter .. " times now" endif Note that this means that filetype plugins don't get a different set of script variables for each buffer. Use local buffer variables instead |b:var|. PREDEFINED VIM VARIABLES *vim-variable* *v:var* *v:* *E963* *E1063* Some variables can be set by the user, but the type cannot be changed. *v:argv* *argv-variable* v:argv The command line arguments Vim was invoked with. This is a list of strings. The first item is the Vim command. See |v:progpath| for the command with full path. *v:beval_col* *beval_col-variable* v:beval_col The number of the column, over which the mouse pointer is. This is the byte index in the |v:beval_lnum| line. Only valid while evaluating the 'balloonexpr' option. *v:beval_bufnr* *beval_bufnr-variable* v:beval_bufnr The number of the buffer, over which the mouse pointer is. Only valid while evaluating the 'balloonexpr' option. *v:beval_lnum* *beval_lnum-variable* v:beval_lnum The number of the line, over which the mouse pointer is. Only valid while evaluating the 'balloonexpr' option. *v:beval_text* *beval_text-variable* v:beval_text The text under or after the mouse pointer. Usually a word as it is useful for debugging a C program. 'iskeyword' applies, but a dot and "->" before the position is included. When on a ']' the text before it is used, including the matching '[' and word before it. When on a Visual area within one line the highlighted text is used. Also see ||. Only valid while evaluating the 'balloonexpr' option. *v:beval_winnr* *beval_winnr-variable* v:beval_winnr The number of the window, over which the mouse pointer is. Only valid while evaluating the 'balloonexpr' option. The first window has number zero (unlike most other places where a window gets a number). *v:beval_winid* *beval_winid-variable* v:beval_winid The |window-ID| of the window, over which the mouse pointer is. Otherwise like v:beval_winnr. *v:char* *char-variable* v:char Argument for evaluating 'formatexpr' and used for the typed character when using in an abbreviation |:map-|. It is also used by the |InsertCharPre| and |InsertEnter| events. *v:charconvert_from* *charconvert_from-variable* v:charconvert_from The name of the character encoding of a file to be converted. Only valid while evaluating the 'charconvert' option. *v:charconvert_to* *charconvert_to-variable* v:charconvert_to The name of the character encoding of a file after conversion. Only valid while evaluating the 'charconvert' option. *v:cmdarg* *cmdarg-variable* v:cmdarg This variable is used for two purposes: 1. The extra arguments given to a file read/write command. Currently these are "++enc=" and "++ff=". This variable is set before an autocommand event for a file read/write command is triggered. There is a leading space to make it possible to append this variable directly after the read/write command. Note: The "+cmd" argument isn't included here, because it will be executed anyway. 2. When printing a PostScript file with ":hardcopy" this is the argument for the ":hardcopy" command. This can be used in 'printexpr'. *v:cmdbang* *cmdbang-variable* v:cmdbang Set like v:cmdarg for a file read/write command. When a "!" was used the value is 1, otherwise it is 0. Note that this can only be used in autocommands. For user commands || can be used. *v:collate* *collate-variable* v:collate The current locale setting for collation order of the runtime environment. This allows Vim scripts to be aware of the current locale encoding. Technical: it's the value of LC_COLLATE. When not using a locale the value is "C". This variable can not be set directly, use the |:language| command. See |multi-lang|. *v:colornames* v:colornames A dictionary that maps color names to hex color strings. These color names can be used with the |highlight-guifg|, |highlight-guibg|, and |highlight-guisp| parameters. The key values in the dictionary (the color names) should be lower cased, because Vim looks up a color by its lower case name. Updating an entry in v:colornames has no immediate effect on the syntax highlighting. The highlight commands (probably in a colorscheme script) need to be re-evaluated in order to use the updated color values. For example: > :let v:colornames['fuscia'] = '#cf3ab4' :let v:colornames['mauve'] = '#915f6d' :highlight Normal guifg=fuscia guibg=mauve < This cannot be used to override the |cterm-colors| but it can be used to override other colors. For example, the X11 colors defined in the `colors/lists/default.vim` (previously defined in |rgb.txt|). When defining new color names in a plugin, the recommended practice is to set a color entry only when it does not already exist. For example: > :call extend(v:colornames, { \ 'fuscia': '#cf3ab4', \ 'mauve': '#915f6d, \ }, 'keep') < Using |extend()| with the 'keep' option updates each color only if it did not exist in |v:colornames|. Doing so allows the user to choose the precise color value for a common name by setting it in their |.vimrc|. It is possible to remove entries from this dictionary but doing so is NOT recommended, because it is disruptive to other scripts. It is also unlikely to achieve the desired result because the |:colorscheme| and |:highlight| commands will both automatically load all `colors/lists/default.vim` color scripts. You can make changes to that file, but make sure to add new keys instead of updating existing ones, otherwise Vim will skip loading the file (thinking is hasn't been changed). *v:completed_item* *completed_item-variable* v:completed_item |Dictionary| containing the |complete-items| for the most recently completed word after |CompleteDone|. The |Dictionary| is empty if the completion failed. Note: Plugins can modify the value to emulate the builtin |CompleteDone| event behavior. *v:count* *count-variable* v:count The count given for the last Normal mode command. Can be used to get the count before a mapping. Read-only. Example: > :map _x :echo "the count is " .. v:count < Note: The is required to remove the line range that you get when typing ':' after a count. When there are two counts, as in "3d2w", they are multiplied, just like what happens in the command, "d6w" for the example. Also used for evaluating the 'formatexpr' option. "count" also works, for backwards compatibility, unless |scriptversion| is 3 or higher. *v:count1* *count1-variable* v:count1 Just like "v:count", but defaults to one when no count is used. *v:ctype* *ctype-variable* v:ctype The current locale setting for characters of the runtime environment. This allows Vim scripts to be aware of the current locale encoding. Technical: it's the value of LC_CTYPE. When not using a locale the value is "C". This variable can not be set directly, use the |:language| command. See |multi-lang|. *v:dying* *dying-variable* v:dying Normally zero. When a deadly signal is caught it's set to one. When multiple signals are caught the number increases. Can be used in an autocommand to check if Vim didn't terminate normally. {only works on Unix} Example: > :au VimLeave * if v:dying | echo "\nAAAAaaaarrrggghhhh!!!\n" | endif < Note: if another deadly signal is caught when v:dying is one, VimLeave autocommands will not be executed. *v:exiting* *exiting-variable* v:exiting Vim exit code. Normally zero, non-zero when something went wrong. The value is v:null before invoking the |VimLeavePre| and |VimLeave| autocmds. See |:q|, |:x| and |:cquit|. Example: > :au VimLeave * echo "Exit value is " .. v:exiting < *v:echospace* *echospace-variable* v:echospace Number of screen cells that can be used for an `:echo` message in the last screen line before causing the |hit-enter-prompt|. Depends on 'showcmd', 'ruler' and 'columns'. You need to check 'cmdheight' for whether there are full-width lines available above the last line. *v:errmsg* *errmsg-variable* v:errmsg Last given error message. It's allowed to set this variable. Example: > :let v:errmsg = "" :silent! next :if v:errmsg != "" : ... handle error < "errmsg" also works, for backwards compatibility, unless |scriptversion| is 3 or higher. *v:errors* *errors-variable* *assert-return* v:errors Errors found by assert functions, such as |assert_true()|. This is a list of strings. The assert functions append an item when an assert fails. The return value indicates this: a one is returned if an item was added to v:errors, otherwise zero is returned. To remove old results make it empty: > :let v:errors = [] < If v:errors is set to anything but a list it is made an empty list by the assert function. *v:event* *event-variable* v:event Dictionary containing information about the current |autocommand|. See the specific event for what it puts in this dictionary. The dictionary is emptied when the |autocommand| finishes, please refer to |dict-identity| for how to get an independent copy of it. Use |deepcopy()| if you want to keep the information after the event triggers. Example: > au TextYankPost * let g:foo = deepcopy(v:event) < *v:exception* *exception-variable* v:exception The value of the exception most recently caught and not finished. See also |v:throwpoint| and |throw-variables|. Example: > :try : throw "oops" :catch /.*/ : echo "caught " .. v:exception :endtry < Output: "caught oops". *v:false* *false-variable* v:false A Number with value zero. Used to put "false" in JSON. See |json_encode()|. When used as a string this evaluates to "v:false". > echo v:false < v:false ~ That is so that eval() can parse the string back to the same value. Read-only. In |Vim9| script "false" can be used which has a boolean type. *v:fcs_reason* *fcs_reason-variable* v:fcs_reason The reason why the |FileChangedShell| event was triggered. Can be used in an autocommand to decide what to do and/or what to set v:fcs_choice to. Possible values: deleted file no longer exists conflict file contents, mode or timestamp was changed and buffer is modified changed file contents has changed mode mode of file changed time only file timestamp changed *v:fcs_choice* *fcs_choice-variable* v:fcs_choice What should happen after a |FileChangedShell| event was triggered. Can be used in an autocommand to tell Vim what to do with the affected buffer: reload Reload the buffer (does not work if the file was deleted). edit Reload the buffer and detect the values for options such as 'fileformat', 'fileencoding', 'binary' (does not work if the file was deleted). ask Ask the user what to do, as if there was no autocommand. Except that when only the timestamp changed nothing will happen. Nothing, the autocommand should do everything that needs to be done. The default is empty. If another (invalid) value is used then Vim behaves like it is empty, there is no warning message. *v:fname* *fname-variable* v:fname When evaluating 'includeexpr': the file name that was detected. Empty otherwise. *v:fname_in* *fname_in-variable* v:fname_in The name of the input file. Valid while evaluating: option used for ~ 'charconvert' file to be converted 'diffexpr' original file 'patchexpr' original file 'printexpr' file to be printed And set to the swap file name for |SwapExists|. *v:fname_out* *fname_out-variable* v:fname_out The name of the output file. Only valid while evaluating: option used for ~ 'charconvert' resulting converted file (*) 'diffexpr' output of diff 'patchexpr' resulting patched file (*) When doing conversion for a write command (e.g., ":w file") it will be equal to v:fname_in. When doing conversion for a read command (e.g., ":e file") it will be a temporary file and different from v:fname_in. *v:fname_new* *fname_new-variable* v:fname_new The name of the new version of the file. Only valid while evaluating 'diffexpr'. *v:fname_diff* *fname_diff-variable* v:fname_diff The name of the diff (patch) file. Only valid while evaluating 'patchexpr'. *v:folddashes* *folddashes-variable* v:folddashes Used for 'foldtext': dashes representing foldlevel of a closed fold. Read-only in the |sandbox|. |fold-foldtext| *v:foldlevel* *foldlevel-variable* v:foldlevel Used for 'foldtext': foldlevel of closed fold. Read-only in the |sandbox|. |fold-foldtext| *v:foldend* *foldend-variable* v:foldend Used for 'foldtext': last line of closed fold. Read-only in the |sandbox|. |fold-foldtext| *v:foldstart* *foldstart-variable* v:foldstart Used for 'foldtext': first line of closed fold. Read-only in the |sandbox|. |fold-foldtext| *v:hlsearch* *hlsearch-variable* v:hlsearch Variable that indicates whether search highlighting is on. Setting it makes sense only if 'hlsearch' is enabled which requires |+extra_search|. Setting this variable to zero acts like the |:nohlsearch| command, setting it to one acts like > let &hlsearch = &hlsearch < Note that the value is restored when returning from a function. |function-search-undo|. *v:insertmode* *insertmode-variable* v:insertmode Used for the |InsertEnter| and |InsertChange| autocommand events. Values: i Insert mode r Replace mode v Virtual Replace mode *v:key* *key-variable* v:key Key of the current item of a |Dictionary|. Only valid while evaluating the expression used with |map()| and |filter()|. Read-only. *v:lang* *lang-variable* v:lang The current locale setting for messages of the runtime environment. This allows Vim scripts to be aware of the current language. Technical: it's the value of LC_MESSAGES. The value is system dependent. This variable can not be set directly, use the |:language| command. It can be different from |v:ctype| when messages are desired in a different language than what is used for character encoding. See |multi-lang|. *v:lc_time* *lc_time-variable* v:lc_time The current locale setting for time messages of the runtime environment. This allows Vim scripts to be aware of the current language. Technical: it's the value of LC_TIME. This variable can not be set directly, use the |:language| command. See |multi-lang|. *v:lnum* *lnum-variable* v:lnum Line number for the 'foldexpr' |fold-expr|, 'formatexpr' and 'indentexpr' expressions, tab page number for 'guitablabel' and 'guitabtooltip'. Only valid while one of these expressions is being evaluated. Read-only when in the |sandbox|. *v:maxcol* *maxcol-variable* v:maxcol Maximum line length. Depending on where it is used it can be screen columns, characters or bytes. The value currently is 2147483647 on all systems. *v:mouse_win* *mouse_win-variable* v:mouse_win Window number for a mouse click obtained with |getchar()|. First window has number 1, like with |winnr()|. The value is zero when there was no mouse button click. *v:mouse_winid* *mouse_winid-variable* v:mouse_winid Window ID for a mouse click obtained with |getchar()|. The value is zero when there was no mouse button click. *v:mouse_lnum* *mouse_lnum-variable* v:mouse_lnum Line number for a mouse click obtained with |getchar()|. This is the text line number, not the screen line number. The value is zero when there was no mouse button click. *v:mouse_col* *mouse_col-variable* v:mouse_col Column number for a mouse click obtained with |getchar()|. This is the screen column number, like with |virtcol()|. The value is zero when there was no mouse button click. *v:none* *none-variable* *None* v:none An empty String. Used to put an empty item in JSON. See |json_encode()|. This can also be used as a function argument to use the default value, see |none-function_argument|. When used as a number this evaluates to zero. When used as a string this evaluates to "v:none". > echo v:none < v:none ~ That is so that eval() can parse the string back to the same value. Read-only. Note that using `== v:none` and `!= v:none` will often give an error. Instead, use `is v:none` and `isnot v:none` . *v:null* *null-variable* v:null An empty String. Used to put "null" in JSON. See |json_encode()|. When used as a number this evaluates to zero. When used as a string this evaluates to "v:null". > echo v:null < v:null ~ That is so that eval() can parse the string back to the same value. Read-only. In |Vim9| script `null` can be used without "v:". In some places `v:null` and `null` can be used for a List, Dict, Job, etc. that is not set. That is slightly different than an empty List, Dict, etc. *v:numbermax* *numbermax-variable* v:numbermax Maximum value of a number. *v:numbermin* *numbermin-variable* v:numbermin Minimum value of a number (negative). *v:numbersize* *numbersize-variable* v:numbersize Number of bits in a Number. This is normally 64, but on some systems it may be 32. *v:oldfiles* *oldfiles-variable* v:oldfiles List of file names that is loaded from the |viminfo| file on startup. These are the files that Vim remembers marks for. The length of the List is limited by the ' argument of the 'viminfo' option (default is 100). When the |viminfo| file is not used the List is empty. Also see |:oldfiles| and |c_#<|. The List can be modified, but this has no effect on what is stored in the |viminfo| file later. If you use values other than String this will cause trouble. {only when compiled with the |+viminfo| feature} *v:option_new* v:option_new New value of the option. Valid while executing an |OptionSet| autocommand. *v:option_old* v:option_old Old value of the option. Valid while executing an |OptionSet| autocommand. Depending on the command used for setting and the kind of option this is either the local old value or the global old value. *v:option_oldlocal* v:option_oldlocal Old local value of the option. Valid while executing an |OptionSet| autocommand. *v:option_oldglobal* v:option_oldglobal Old global value of the option. Valid while executing an |OptionSet| autocommand. *v:option_type* v:option_type Scope of the set command. Valid while executing an |OptionSet| autocommand. Can be either "global" or "local" *v:option_command* v:option_command Command used to set the option. Valid while executing an |OptionSet| autocommand. value option was set via ~ "setlocal" |:setlocal| or ":let l:xxx" "setglobal" |:setglobal| or ":let g:xxx" "set" |:set| or |:let| "modeline" |modeline| *v:operator* *operator-variable* v:operator The last operator given in Normal mode. This is a single character except for commands starting with or , in which case it is two characters. Best used alongside |v:prevcount| and |v:register|. Useful if you want to cancel Operator-pending mode and then use the operator, e.g.: > :omap O :call MyMotion(v:operator) < The value remains set until another operator is entered, thus don't expect it to be empty. v:operator is not set for |:delete|, |:yank| or other Ex commands. Read-only. *v:prevcount* *prevcount-variable* v:prevcount The count given for the last but one Normal mode command. This is the v:count value of the previous command. Useful if you want to cancel Visual or Operator-pending mode and then use the count, e.g.: > :vmap % :call MyFilter(v:prevcount) < Read-only. *v:profiling* *profiling-variable* v:profiling Normally zero. Set to one after using ":profile start". See |profiling|. *v:progname* *progname-variable* v:progname Contains the name (with path removed) with which Vim was invoked. Allows you to do special initialisations for |view|, |evim| etc., or any other name you might symlink to Vim. Read-only. *v:progpath* *progpath-variable* v:progpath Contains the command with which Vim was invoked, in a form that when passed to the shell will run the same Vim executable as the current one (if $PATH remains unchanged). Useful if you want to message a Vim server using a |--remote-expr|. To get the full path use: > echo exepath(v:progpath) < If the command has a relative path it will be expanded to the full path, so that it still works after `:cd`. Thus starting "./vim" results in "/home/user/path/to/vim/src/vim". On Linux and other systems it will always be the full path. On Mac it may just be "vim" and using exepath() as mentioned above should be used to get the full path. On MS-Windows the executable may be called "vim.exe", but the ".exe" is not added to v:progpath. Read-only. *v:python3_version* *python3-version-variable* v:python3_version Version of Python 3 that Vim was built against. When Python is loaded dynamically (|python-dynamic|), this version should exactly match the Python library up to the minor version (e.g. 3.10.2 and 3.10.3 are compatible as the minor version is "10", whereas 3.9.4 and 3.10.3 are not compatible). When |python-stable-abi| is used, this will be the minimum Python version that you can use instead. (e.g. if v:python3_version indicates 3.9, you can use 3.9, 3.10, or anything above). This number is encoded as a hex number following Python ABI versioning conventions. Do the following to have a human-readable full version in hex: > echo printf("%08X", v:python3_version) < You can obtain only the minor version by doing: > echo and(v:python3_version>>16,0xff) < Read-only. *v:register* *register-variable* v:register The name of the register in effect for the current normal mode command (regardless of whether that command actually used a register). Or for the currently executing normal mode mapping (use this in custom commands that take a register). If none is supplied it is the default register '"', unless 'clipboard' contains "unnamed" or "unnamedplus", then it is '*' or '+'. Also see |getreg()| and |setreg()| *v:scrollstart* *scrollstart-variable* v:scrollstart String describing the script or function that caused the screen to scroll up. It's only set when it is empty, thus the first reason is remembered. It is set to "Unknown" for a typed command. This can be used to find out why your script causes the hit-enter prompt. *v:servername* *servername-variable* v:servername The resulting registered |client-server-name| if any. Read-only. v:searchforward *v:searchforward* *searchforward-variable* Search direction: 1 after a forward search, 0 after a backward search. It is reset to forward when directly setting the last search pattern, see |quote/|. Note that the value is restored when returning from a function. |function-search-undo|. Read-write. *v:shell_error* *shell_error-variable* v:shell_error Result of the last shell command. When non-zero, the last shell command had an error. When zero, there was no problem. This only works when the shell returns the error code to Vim. The value -1 is often used when the command could not be executed. Read-only. Example: > :!mv foo bar :if v:shell_error : echo 'could not rename "foo" to "bar"!' :endif < "shell_error" also works, for backwards compatibility, unless |scriptversion| is 3 or higher. *v:sizeofint* *sizeofint-variable* v:sizeofint Number of bytes in an int. Depends on how Vim was compiled. This is only useful for deciding whether a test will give the expected result. *v:sizeoflong* *sizeoflong-variable* v:sizeoflong Number of bytes in a long. Depends on how Vim was compiled. This is only useful for deciding whether a test will give the expected result. *v:sizeofpointer* *sizeofpointer-variable* v:sizeofpointer Number of bytes in a pointer. Depends on how Vim was compiled. This is only useful for deciding whether a test will give the expected result. *v:statusmsg* *statusmsg-variable* v:statusmsg Last given status message. It's allowed to set this variable. *v:swapname* *swapname-variable* v:swapname Only valid when executing |SwapExists| autocommands: Name of the swap file found. Read-only. *v:swapchoice* *swapchoice-variable* v:swapchoice |SwapExists| autocommands can set this to the selected choice for handling an existing swap file: 'o' Open read-only 'e' Edit anyway 'r' Recover 'd' Delete swapfile 'q' Quit 'a' Abort The value should be a single-character string. An empty value results in the user being asked, as would happen when there is no SwapExists autocommand. The default is empty. *v:swapcommand* *swapcommand-variable* v:swapcommand Normal mode command to be executed after a file has been opened. Can be used for a |SwapExists| autocommand to have another Vim open the file and jump to the right place. For example, when jumping to a tag the value is ":tag tagname\r". For ":edit +cmd file" the value is ":cmd\r". *v:t_TYPE* *v:t_bool* *t_bool-variable* v:t_bool Value of |Boolean| type. Read-only. See: |type()| *v:t_channel* *t_channel-variable* v:t_channel Value of |Channel| type. Read-only. See: |type()| *v:t_dict* *t_dict-variable* v:t_dict Value of |Dictionary| type. Read-only. See: |type()| *v:t_float* *t_float-variable* v:t_float Value of |Float| type. Read-only. See: |type()| *v:t_func* *t_func-variable* v:t_func Value of |Funcref| type. Read-only. See: |type()| *v:t_job* *t_job-variable* v:t_job Value of |Job| type. Read-only. See: |type()| *v:t_list* *t_list-variable* v:t_list Value of |List| type. Read-only. See: |type()| *v:t_none* *t_none-variable* v:t_none Value of |None| type. Read-only. See: |type()| *v:t_number* *t_number-variable* v:t_number Value of |Number| type. Read-only. See: |type()| *v:t_string* *t_string-variable* v:t_string Value of |String| type. Read-only. See: |type()| *v:t_blob* *t_blob-variable* v:t_blob Value of |Blob| type. Read-only. See: |type()| *v:t_class* *t_class-variable* v:t_class Value of |class| type. Read-only. See: |type()| *v:t_object* *t_object-variable* v:t_object Value of |object| type. Read-only. See: |type()| *v:t_typealias* *t_typealias-variable* v:t_typealias Value of |typealias| type. Read-only. See: |type()| *v:t_enum* *t_enum-variable* v:t_enum Value of |enum| type. Read-only. See: |type()| *v:t_enumvalue* *t_enumvalue-variable* v:t_enumvalue Value of |enumvalue| type. Read-only. See: |type()| *v:termresponse* *termresponse-variable* v:termresponse The escape sequence returned by the terminal for the |t_RV| termcap entry. It is set when Vim receives an escape sequence that starts with ESC [ or CSI, then '>' or '?' and ends in a 'c', with only digits and ';' in between. When this option is set, the TermResponse autocommand event is fired, so that you can react to the response from the terminal. The TermResponseAll event is also fired, with set to "version". You can use |terminalprops()| to see what Vim figured out about the terminal. The response from a new xterm is: "[> Pp ; Pv ; Pc c". Pp is the terminal type: 0 for vt100 and 1 for vt220. Pv is the patch level (since this was introduced in patch 95, it's always 95 or higher). Pc is always zero. If Pv is 141 or higher then Vim will try to request terminal codes. This only works with xterm |xterm-codes|. {only when compiled with |+termresponse| feature} *v:termblinkresp* v:termblinkresp The escape sequence returned by the terminal for the |t_RC| termcap entry. This is used to find out whether the terminal cursor is blinking. This is used by |term_getcursor()|. When this option is set, the TermResponseAll autocommand event is fired, with set to "cursorblink". *v:termstyleresp* v:termstyleresp The escape sequence returned by the terminal for the |t_RS| termcap entry. This is used to find out what the shape of the cursor is. This is used by |term_getcursor()|. When this option is set, the TermResponseAll autocommand event is fired, with set to "cursorshape". *v:termrbgresp* v:termrbgresp The escape sequence returned by the terminal for the |t_RB| termcap entry. This is used to find out what the terminal background color is; see 'background'. When this option is set, the TermResponseAll autocommand event is fired, with set to "background". *v:termrfgresp* v:termrfgresp The escape sequence returned by the terminal for the |t_RF| termcap entry. This is used to find out what the terminal foreground color is. When this option is set, the TermResponseAll autocommand event is fired, with set to "foreground". *v:termu7resp* v:termu7resp The escape sequence returned by the terminal for the |t_u7| termcap entry. This is used to find out what the terminal does with ambiguous width characters, see 'ambiwidth'. When this option is set, the TermResponseAll autocommand event is fired, with set to "ambiguouswidth". *v:testing* *testing-variable* v:testing Must be set before using `test_garbagecollect_now()`. Also, when set certain error messages won't be shown for 2 seconds. (e.g. "'dictionary' option is empty") *v:this_session* *this_session-variable* v:this_session Full filename of the last loaded or saved session file. See |:mksession|. It is allowed to set this variable. When no session file has been saved, this variable is empty. "this_session" also works, for backwards compatibility, unless |scriptversion| is 3 or higher *v:throwpoint* *throwpoint-variable* v:throwpoint The point where the exception most recently caught and not finished was thrown. Not set when commands are typed. See also |v:exception| and |throw-variables|. Example: > :try : throw "oops" :catch /.*/ : echo "Exception from" v:throwpoint :endtry < Output: "Exception from test.vim, line 2" *v:true* *true-variable* v:true A Number with value one. Used to put "true" in JSON. See |json_encode()|. When used as a string this evaluates to "v:true". > echo v:true < v:true ~ That is so that eval() can parse the string back to the same value. Read-only. In |Vim9| script "true" can be used which has a boolean type. *v:val* *val-variable* v:val Value of the current item of a |List| or |Dictionary|. Only valid while evaluating the expression used with |map()| and |filter()|. Read-only. *v:version* *version-variable* v:version Version number of Vim: Major version number times 100 plus minor version number. Version 5.0 is 500. Version 5.1 is 501. Read-only. "version" also works, for backwards compatibility, unless |scriptversion| is 3 or higher. Use |has()| to check if a certain patch was included, e.g.: > if has("patch-7.4.123") < Note that patch numbers are specific to the version, thus both version 5.0 and 5.1 may have a patch 123, but these are completely different. *v:versionlong* *versionlong-variable* v:versionlong Like v:version, but also including the patchlevel in the last four digits. Version 8.1 with patch 123 has value 8010123. This can be used like this: > if v:versionlong >= 8010123 < However, if there are gaps in the list of patches included this will not work well. This can happen if a recent patch was included into an older version, e.g. for a security fix. Use the has() function to make sure the patch is actually included. *v:vim_did_enter* *vim_did_enter-variable* v:vim_did_enter Zero until most of startup is done. It is set to one just before |VimEnter| autocommands are triggered. *v:warningmsg* *warningmsg-variable* v:warningmsg Last given warning message. It's allowed to set this variable. *v:windowid* *windowid-variable* v:windowid When any X11 based GUI is running or when running in a terminal and Vim connects to the X server (|-X|) this will be set to the window ID. When an MS-Windows GUI is running this will be set to the window handle. Otherwise the value is zero. Note: for windows inside Vim use |winnr()| or |win_getid()|, see |window-ID|. ============================================================================== 4. Builtin Functions *functions* See |function-list| for a list grouped by what the function is used for. The alphabetic list of all builtin functions and details are in a separate help file: |builtin-functions|. ============================================================================== 5. Defining functions *user-functions* New functions can be defined. These can be called just like builtin functions. The function takes arguments, executes a sequence of Ex commands and can return a value. You can find most information about defining functions in |userfunc.txt|. For Vim9 functions, which execute much faster, support type checking and more, see |vim9.txt|. ============================================================================== 6. Curly braces names *curly-braces-names* In most places where you can use a variable, you can use a "curly braces name" variable. This is a regular variable name with one or more expressions wrapped in braces {} like this: > my_{adjective}_variable This only works in legacy Vim script, not in |Vim9| script. When Vim encounters this, it evaluates the expression inside the braces, puts that in place of the expression, and re-interprets the whole as a variable name. So in the above example, if the variable "adjective" was set to "noisy", then the reference would be to "my_noisy_variable", whereas if "adjective" was set to "quiet", then it would be to "my_quiet_variable". One application for this is to create a set of variables governed by an option value. For example, the statement > echo my_{&background}_message would output the contents of "my_dark_message" or "my_light_message" depending on the current value of 'background'. You can use multiple brace pairs: > echo my_{adverb}_{adjective}_message ..or even nest them: > echo my_{ad{end_of_word}}_message where "end_of_word" is either "verb" or "jective". However, the expression inside the braces must evaluate to a valid single variable name, e.g. this is invalid: > :let foo='a + b' :echo c{foo}d .. since the result of expansion is "ca + bd", which is not a variable name. *curly-braces-function-names* You can call and define functions by an evaluated name in a similar way. Example: > :let func_end='whizz' :call my_func_{func_end}(parameter) This would call the function "my_func_whizz(parameter)". This does NOT work: > :let i = 3 :let @{i} = '' " error :echo @{i} " error ============================================================================== 7. Commands *expression-commands* Note: in |Vim9| script `:let` is not used. `:var` is used for variable declarations and assignments do not use a command. |vim9-declaration| :let {var-name} = {expr1} *:let* *E18* Set internal variable {var-name} to the result of the expression {expr1}. The variable will get the type from the {expr}. If {var-name} didn't exist yet, it is created. :let {var-name}[{idx}] = {expr1} *E689* *E1141* Set a list item to the result of the expression {expr1}. {var-name} must refer to a list and {idx} must be a valid index in that list. For nested list the index can be repeated. This cannot be used to add an item to a |List|. This cannot be used to set a byte in a String. You can do that like this: > :let var = var[0:2] .. 'X' .. var[4:] < When {var-name} is a |Blob| then {idx} can be the length of the blob, in which case one byte is appended. *E711* *E719* *E1165* *E1166* *E1183* :let {var-name}[{idx1}:{idx2}] = {expr1} *E708* *E709* *E710* Set a sequence of items in a |List| to the result of the expression {expr1}, which must be a list with the correct number of items. {idx1} can be omitted, zero is used instead. {idx2} can be omitted, meaning the end of the list. When the selected range of items is partly past the end of the list, items will be added. *:let+=* *:let-=* *:letstar=* *:let/=* *:let%=* *:let.=* *:let..=* *E734* *E985* *E1019* :let {var} += {expr1} Like ":let {var} = {var} + {expr1}". :let {var} -= {expr1} Like ":let {var} = {var} - {expr1}". :let {var} *= {expr1} Like ":let {var} = {var} * {expr1}". :let {var} /= {expr1} Like ":let {var} = {var} / {expr1}". :let {var} %= {expr1} Like ":let {var} = {var} % {expr1}". :let {var} .= {expr1} Like ":let {var} = {var} . {expr1}". :let {var} ..= {expr1} Like ":let {var} = {var} .. {expr1}". These fail if {var} was not set yet and when the type of {var} and {expr1} don't fit the operator. `+=` modifies a |List| or a |Blob| in-place instead of creating a new one. `.=` is not supported with Vim script version 2 and later, see |vimscript-version|. :let ${env-name} = {expr1} *:let-environment* *:let-$* Set environment variable {env-name} to the result of the expression {expr1}. The type is always String. On some systems making an environment variable empty causes it to be deleted. Many systems do not make a difference between an environment variable that is not set and an environment variable that is empty. :let ${env-name} .= {expr1} Append {expr1} to the environment variable {env-name}. If the environment variable didn't exist yet this works like "=". :let @{reg-name} = {expr1} *:let-register* *:let-@* Write the result of the expression {expr1} in register {reg-name}. {reg-name} must be a single letter, and must be the name of a writable register (see |registers|). "@@" can be used for the unnamed register, "@/" for the search pattern. If the result of {expr1} ends in a or , the register will be linewise, otherwise it will be set to characterwise. This can be used to clear the last search pattern: > :let @/ = "" < This is different from searching for an empty string, that would match everywhere. :let @{reg-name} .= {expr1} Append {expr1} to register {reg-name}. If the register was empty it's like setting it to {expr1}. :let &{option-name} = {expr1} *:let-option* *:let-&* Set option {option-name} to the result of the expression {expr1}. A String or Number value is always converted to the type of the option. For an option local to a window or buffer the effect is just like using the |:set| command: both the local value and the global value are changed. Example: > :let &path = &path .. ',/usr/local/include' < This also works for terminal codes in the form t_xx. But only for alphanumerical names. Example: > :let &t_k1 = "\[234;" < When the code does not exist yet it will be created as a terminal key code, there is no error. :let &{option-name} .= {expr1} For a string option: Append {expr1} to the value. Does not insert a comma like |:set+=|. :let &{option-name} += {expr1} :let &{option-name} -= {expr1} For a number or boolean option: Add or subtract {expr1}. :let &l:{option-name} = {expr1} :let &l:{option-name} .= {expr1} :let &l:{option-name} += {expr1} :let &l:{option-name} -= {expr1} Like above, but only set the local value of an option (if there is one). Works like |:setlocal|. :let &g:{option-name} = {expr1} :let &g:{option-name} .= {expr1} :let &g:{option-name} += {expr1} :let &g:{option-name} -= {expr1} Like above, but only set the global value of an option (if there is one). Works like |:setglobal|. *E1093* :let [{name1}, {name2}, ...] = {expr1} *:let-unpack* *E687* *E688* {expr1} must evaluate to a |List|. The first item in the list is assigned to {name1}, the second item to {name2}, etc. The number of names must match the number of items in the |List|. Each name can be one of the items of the ":let" command as mentioned above. Example: > :let [s, item] = GetItem(s) < Detail: {expr1} is evaluated first, then the assignments are done in sequence. This matters if {name2} depends on {name1}. Example: > :let x = [0, 1] :let i = 0 :let [i, x[i]] = [1, 2] :echo x < The result is [0, 2]. :let [{name1}, {name2}, ...] .= {expr1} :let [{name1}, {name2}, ...] += {expr1} :let [{name1}, {name2}, ...] -= {expr1} Like above, but append/add/subtract the value for each |List| item. :let [{name}, ..., ; {lastname}] = {expr1} *E452* Like |:let-unpack| above, but the |List| may have more items than there are names. A list of the remaining items is assigned to {lastname}. If there are no remaining items {lastname} is set to an empty list. Example: > :let [a, b; rest] = ["aval", "bval", 3, 4] < :let [{name}, ..., ; {lastname}] .= {expr1} :let [{name}, ..., ; {lastname}] += {expr1} :let [{name}, ..., ; {lastname}] -= {expr1} Like above, but append/add/subtract the value for each |List| item. *:let=<<* *:let-heredoc* *E990* *E991* *E172* *E221* *E1145* :let {var-name} =<< [trim] [eval] {endmarker} text... text... {endmarker} Set internal variable {var-name} to a |List| containing the lines of text bounded by the string {endmarker}. If "eval" is not specified, then each line of text is used as a |literal-string|, except that single quotes does not need to be doubled. If "eval" is specified, then any Vim expression in the form {expr} is evaluated and the result replaces the expression, like with |interpolated-string|. Example where $HOME is expanded: > let lines =<< trim eval END some text See the file {$HOME}/.vimrc more text END < There can be multiple Vim expressions in a single line but an expression cannot span multiple lines. If any expression evaluation fails, then the assignment fails. {endmarker} must not contain white space. {endmarker} cannot start with a lower case character. The last line should end only with the {endmarker} string without any other character. Watch out for white space after {endmarker}! Without "trim" any white space characters in the lines of text are preserved. If "trim" is specified before {endmarker}, then indentation is stripped so you can do: > let text =<< trim END if ok echo 'done' endif END < Results in: ["if ok", " echo 'done'", "endif"] The marker must line up with "let" and the indentation of the first line is removed from all the text lines. Specifically: all the leading indentation exactly matching the leading indentation of the first non-empty text line is stripped from the input lines. All leading indentation exactly matching the leading indentation before `let` is stripped from the line containing {endmarker}. Note that the difference between space and tab matters here. If {var-name} didn't exist yet, it is created. Cannot be followed by another command, but can be followed by a comment. To avoid line continuation to be applied, consider adding 'C' to 'cpoptions': > set cpo+=C let var =<< END \ leading backslash END set cpo-=C < Examples: > let var1 =<< END Sample text 1 Sample text 2 Sample text 3 END let data =<< trim DATA 1 2 3 4 5 6 7 8 DATA let code =<< trim eval CODE let v = {10 + 20} let h = "{$HOME}" let s = "{Str1()} abc {Str2()}" let n = {MyFunc(3, 4)} CODE < *E121* :let {var-name} .. List the value of variable {var-name}. Multiple variable names may be given. Special names recognized here: *E738* g: global variables b: local buffer variables w: local window variables t: local tab page variables s: script-local variables l: local function variables v: Vim variables. This does not work in Vim9 script. |vim9-declaration| :let List the values of all variables. The type of the variable is indicated before the value: String # Number * Funcref This does not work in Vim9 script. |vim9-declaration| :unl[et][!] {name} ... *:unlet* *:unl* *E108* *E795* *E1081* Remove the internal variable {name}. Several variable names can be given, they are all removed. The name may also be a |List| or |Dictionary| item. With [!] no error message is given for non-existing variables. One or more items from a |List| can be removed: > :unlet list[3] " remove fourth item :unlet list[3:] " remove fourth item to last < One item from a |Dictionary| can be removed at a time: > :unlet dict['two'] :unlet dict.two < This is especially useful to clean up used global variables and script-local variables (these are not deleted when the script ends). Function-local variables are automatically deleted when the function ends. In |Vim9| script variables declared in a function or script cannot be removed. :unl[et] ${env-name} ... *:unlet-environment* *:unlet-$* Remove environment variable {env-name}. Can mix {name} and ${env-name} in one :unlet command. No error message is given for a non-existing variable, also without !. If the system does not support deleting an environment variable, it is made empty. *:cons* *:const* *E1018* :cons[t] {var-name} = {expr1} :cons[t] [{name1}, {name2}, ...] = {expr1} :cons[t] [{name}, ..., ; {lastname}] = {expr1} :cons[t] {var-name} =<< [trim] {marker} text... text... {marker} Similar to |:let|, but additionally lock the variable after setting the value. This is the same as locking the variable with |:lockvar| just after |:let|, thus: > :const x = 1 < is equivalent to: > :let x = 1 :lockvar! x < NOTE: in Vim9 script `:const` works differently, see |vim9-const| This is useful if you want to make sure the variable is not modified. If the value is a List or Dictionary literal then the items also cannot be changed: > const ll = [1, 2, 3] let ll[1] = 5 " Error! < Nested references are not locked: > let lvar = ['a'] const lconst = [0, lvar] let lconst[0] = 2 " Error! let lconst[1][0] = 'b' " OK < *E995* |:const| does not allow to for changing a variable: > :let x = 1 :const x = 2 " Error! < *E996* Note that environment variables, option values and register values cannot be used here, since they cannot be locked. :cons[t] :cons[t] {var-name} If no argument is given or only {var-name} is given, the behavior is the same as |:let|. :lockv[ar][!] [depth] {name} ... *:lockvar* *:lockv* Lock the internal variable {name}. Locking means that it can no longer be changed (until it is unlocked). A locked variable can be deleted: > :lockvar v :let v = 'asdf' " fails! :unlet v " works < *E741* *E940* *E1118* *E1119* *E1120* *E1121* *E1122* If you try to change a locked variable you get an error message: "E741: Value is locked: {name}". If you try to lock or unlock a built-in variable you get an error message: "E940: Cannot lock or unlock variable {name}". [depth] is relevant when locking a |List| or |Dictionary|. It specifies how deep the locking goes: 0 Lock the variable {name} but not its value. 1 Lock the |List| or |Dictionary| itself, cannot add or remove items, but can still change their values. 2 Also lock the values, cannot change the items. If an item is a |List| or |Dictionary|, cannot add or remove items, but can still change the values. 3 Like 2 but for the |List| / |Dictionary| in the |List| / |Dictionary|, one level deeper. The default [depth] is 2, thus when {name} is a |List| or |Dictionary| the values cannot be changed. Example with [depth] 0: > let mylist = [1, 2, 3] lockvar 0 mylist let mylist[0] = 77 " OK call add(mylist, 4) " OK let mylist = [7, 8, 9] " Error! < *E743* For unlimited depth use [!] and omit [depth]. However, there is a maximum depth of 100 to catch loops. Note that when two variables refer to the same |List| and you lock one of them, the |List| will also be locked when used through the other variable. Example: > :let l = [0, 1, 2, 3] :let cl = l :lockvar l :let cl[1] = 99 " won't work! < You may want to make a copy of a list to avoid this. See |deepcopy()|. *E1391* *E1392* Locking and unlocking object and class variables is currently NOT supported. :unlo[ckvar][!] [depth] {name} ... *:unlockvar* *:unlo* *E1246* Unlock the internal variable {name}. Does the opposite of |:lockvar|. If {name} does not exist: - In |Vim9| script an error is given. - In legacy script this is silently ignored. :if {expr1} *:if* *:end* *:endif* *:en* *E171* *E579* *E580* :en[dif] Execute the commands until the next matching `:else` or `:endif` if {expr1} evaluates to non-zero. Although the short forms work, it is recommended to always use `:endif` to avoid confusion and to make auto-indenting work properly. From Vim version 4.5 until 5.0, every Ex command in between the `:if` and `:endif` is ignored. These two commands were just to allow for future expansions in a backward compatible way. Nesting was allowed. Note that any `:else` or `:elseif` was ignored, the `else` part was not executed either. You can use this to remain compatible with older versions: > :if version >= 500 : version-5-specific-commands :endif < The commands still need to be parsed to find the `endif`. Sometimes an older Vim has a problem with a new command. For example, `:silent` is recognized as a `:substitute` command. In that case `:execute` can avoid problems: > :if version >= 600 : execute "silent 1,$delete" :endif < In |Vim9| script `:endif` cannot be shortened, to improve script readability. NOTE: The `:append` and `:insert` commands don't work properly in between `:if` and `:endif`. *:else* *:el* *E581* *E583* :el[se] Execute the commands until the next matching `:else` or `:endif` if they previously were not being executed. In |Vim9| script `:else` cannot be shortened, to improve script readability. *:elseif* *:elsei* *E582* *E584* :elsei[f] {expr1} Short for `:else` `:if`, with the addition that there is no extra `:endif`. In |Vim9| script `:elseif` cannot be shortened, to improve script readability. :wh[ile] {expr1} *:while* *:endwhile* *:wh* *:endw* *E170* *E585* *E588* *E733* :endw[hile] Repeat the commands between `:while` and `:endwhile`, as long as {expr1} evaluates to non-zero. When an error is detected from a command inside the loop, execution continues after the `endwhile`. Example: > :let lnum = 1 :while lnum <= line("$") :call FixLine(lnum) :let lnum = lnum + 1 :endwhile < In |Vim9| script `:while` and `:endwhile` cannot be shortened, to improve script readability. NOTE: The `:append` and `:insert` commands don't work properly inside a `:while` and `:for` loop. :for {var} in {object} *:for* *E690* *E732* :endfo[r] *:endfo* *:endfor* Repeat the commands between `:for` and `:endfor` for each item in {object}. {object} can be a |List|, a |Blob| or a |String|. *E1177* Variable {var} is set to the value of each item. In |Vim9| script the loop variable must not have been declared yet, unless when it is a global/window/tab/buffer variable. When an error is detected for a command inside the loop, execution continues after the `endfor`. Changing {object} inside the loop affects what items are used. Make a copy if this is unwanted: > :for item in copy(mylist) < When {object} is a |List| and not making a copy, in legacy script Vim stores a reference to the next item in the |List| before executing the commands with the current item. Thus the current item can be removed without effect. Removing any later item means it will not be found. Thus the following example works (an inefficient way to make a |List| empty): > for item in mylist call remove(mylist, 0) endfor < Note that reordering the |List| (e.g., with sort() or reverse()) may have unexpected effects. In |Vim9| script the index is used. If an item before the current one is deleted the next item will be skipped. When {object} is a |Blob|, Vim always makes a copy to iterate over. Unlike with |List|, modifying the |Blob| does not affect the iteration. When {object} is a |String| each item is a string with one character, plus any combining characters. In |Vim9| script `:endfor` cannot be shortened, to improve script readability. :for [{var1}, {var2}, ...] in {listlist} :endfo[r] *E1140* Like `:for` above, but each item in {listlist} must be a list, of which each item is assigned to {var1}, {var2}, etc. Example: > :for [lnum, col] in [[1, 3], [2, 5], [3, 8]] :echo getline(lnum)[col] :endfor < *:continue* *:con* *E586* :con[tinue] When used inside a `:while` or `:for` loop, jumps back to the start of the loop. If it is used after a `:try` inside the loop but before the matching `:finally` (if present), the commands following the `:finally` up to the matching `:endtry` are executed first. This process applies to all nested `:try`s inside the loop. The outermost `:endtry` then jumps back to the start of the loop. In |Vim9| script `:cont` is the shortest form, to improve script readability. *:break* *:brea* *E587* :brea[k] When used inside a `:while` or `:for` loop, skips to the command after the matching `:endwhile` or `:endfor`. If it is used after a `:try` inside the loop but before the matching `:finally` (if present), the commands following the `:finally` up to the matching `:endtry` are executed first. This process applies to all nested `:try`s inside the loop. The outermost `:endtry` then jumps to the command after the loop. In |Vim9| script `:break` cannot be shortened, to improve script readability. :try *:try* *:endt* *:endtry* *E600* *E601* *E602* *E1032* :endt[ry] Change the error handling for the commands between `:try` and `:endtry` including everything being executed across `:source` commands, function calls, or autocommand invocations. When an error or interrupt is detected and there is a `:finally` command following, execution continues after the `:finally`. Otherwise, or when the `:endtry` is reached thereafter, the next (dynamically) surrounding `:try` is checked for a corresponding `:finally` etc. Then the script processing is terminated. Whether a function definition has an "abort" argument does not matter. Example: > try | call Unknown() | finally | echomsg "cleanup" | endtry echomsg "not reached" < Moreover, an error or interrupt (dynamically) inside `:try` and `:endtry` is converted to an exception. It can be caught as if it were thrown by a `:throw` command (see `:catch`). In this case, the script processing is not terminated. The value "Vim:Interrupt" is used for an interrupt exception. An error in a Vim command is converted to a value of the form "Vim({command}):{errmsg}", other errors are converted to a value of the form "Vim:{errmsg}". {command} is the full command name, and {errmsg} is the message that is displayed if the error exception is not caught, always beginning with the error number. Examples: > try | sleep 100 | catch /^Vim:Interrupt$/ | endtry try | edit | catch /^Vim(edit):E\d\+/ | echo "error" | endtry < In |Vim9| script `:endtry` cannot be shortened, to improve script readability. *:cat* *:catch* *E603* *E604* *E605* *E654* *E1033* :cat[ch] /{pattern}/ The following commands until the next `:catch`, `:finally`, or `:endtry` that belongs to the same `:try` as the `:catch` are executed when an exception matching {pattern} is being thrown and has not yet been caught by a previous `:catch`. Otherwise, these commands are skipped. When {pattern} is omitted all errors are caught. Examples: > :catch /^Vim:Interrupt$/ " catch interrupts (CTRL-C) :catch /^Vim\%((\a\+)\)\=:E/ " catch all Vim errors :catch /^Vim\%((\a\+)\)\=:/ " catch errors and interrupts :catch /^Vim(write):/ " catch all errors in :write :catch /^Vim\%((\a\+)\)\=:E123:/ " catch error E123 :catch /my-exception/ " catch user exception :catch /.*/ " catch everything :catch " same as /.*/ < Another character can be used instead of / around the {pattern}, so long as it does not have a special meaning (e.g., '|' or '"') and doesn't occur inside {pattern}. *E1067* Information about the exception is available in |v:exception|. Also see |throw-variables|. NOTE: It is not reliable to ":catch" the TEXT of an error message because it may vary in different locales. In |Vim9| script `:catch` cannot be shortened, to improve script readability. *:fina* *:finally* *E606* *E607* :fina[lly] The following commands until the matching `:endtry` are executed whenever the part between the matching `:try` and the `:finally` is left: either by falling through to the `:finally` or by a `:continue`, `:break`, `:finish`, or `:return`, or by an error or interrupt or exception (see `:throw`). In |Vim9| script `:finally` cannot be shortened, to improve script readability and avoid confusion with `:final`. *:th* *:throw* *E608* *E1129* :th[row] {expr1} The {expr1} is evaluated and thrown as an exception. If the ":throw" is used after a `:try` but before the first corresponding `:catch`, commands are skipped until the first `:catch` matching {expr1} is reached. If there is no such `:catch` or if the ":throw" is used after a `:catch` but before the `:finally`, the commands following the `:finally` (if present) up to the matching `:endtry` are executed. If the `:throw` is after the `:finally`, commands up to the `:endtry` are skipped. At the ":endtry", this process applies again for the next dynamically surrounding `:try` (which may be found in a calling function or sourcing script), until a matching `:catch` has been found. If the exception is not caught, the command processing is terminated. Example: > :try | throw "oops" | catch /^oo/ | echo "caught" | endtry < Note that "catch" may need to be on a separate line for when an error causes the parsing to skip the whole line and not see the "|" that separates the commands. In |Vim9| script `:throw` cannot be shortened, to improve script readability. *:ec* *:echo* :ec[ho] {expr1} .. Echoes each {expr1}, with a space in between. The first {expr1} starts on a new line. Also see |:comment|. Use "\n" to start a new line. Use "\r" to move the cursor to the first column. Uses the highlighting set by the `:echohl` command. Cannot be followed by a comment. Example: > :echo "the value of 'shell' is" &shell < *:echo-redraw* A later redraw may make the message disappear again. And since Vim mostly postpones redrawing until it's finished with a sequence of commands this happens quite often. To avoid that a command from before the `:echo` causes a redraw afterwards (redraws are often postponed until you type something), force a redraw with the `:redraw` command. Example: > :new | redraw | echo "there is a new window" < *:echon* :echon {expr1} .. Echoes each {expr1}, without anything added. Also see |:comment|. Uses the highlighting set by the `:echohl` command. Cannot be followed by a comment. Example: > :echon "the value of 'shell' is " &shell < Note the difference between using `:echo`, which is a Vim command, and `:!echo`, which is an external shell command: > :!echo % --> filename < The arguments of ":!" are expanded, see |:_%|. > :!echo "%" --> filename or "filename" < Like the previous example. Whether you see the double quotes or not depends on your 'shell'. > :echo % --> nothing < The '%' is an illegal character in an expression. > :echo "%" --> % < This just echoes the '%' character. > :echo expand("%") --> filename < This calls the expand() function to expand the '%'. *:echoh* *:echohl* :echoh[l] {name} Use the highlight group {name} for the following `:echo`, `:echon` and `:echomsg` commands. Also used for the `input()` prompt. Example: > :echohl WarningMsg | echo "Don't panic!" | echohl None < Don't forget to set the group back to "None", otherwise all following echo's will be highlighted. *:echom* *:echomsg* :echom[sg] {expr1} .. Echo the expression(s) as a true message, saving the message in the |message-history|. Spaces are placed between the arguments as with the `:echo` command. But unprintable characters are displayed, not interpreted. The parsing works slightly different from `:echo`, more like `:execute`. All the expressions are first evaluated and concatenated before echoing anything. If expressions does not evaluate to a Number or String, string() is used to turn it into a string. Uses the highlighting set by the `:echohl` command. Example: > :echomsg "It's a Zizzer Zazzer Zuzz, as you can plainly see." < See |:echo-redraw| to avoid the message disappearing when the screen is redrawn. *:echow* *:echowin* *:echowindow* :[N]echow[indow] {expr1} .. Like |:echomsg| but when the messages popup window is available the message is displayed there. This means it will show for three seconds and avoid a |hit-enter| prompt. If you want to hide it before that, press Esc in Normal mode (when it would otherwise beep). If it disappears too soon you can use `:messages` to see the text. When [N] is given then the window will show up for this number of seconds. The last `:echowindow` with a count matters, it is used once only. The message window is available when Vim was compiled with the +timer and the +popupwin features. *:echoe* *:echoerr* :echoe[rr] {expr1} .. Echo the expression(s) as an error message, saving the message in the |message-history|. When used in a script or function the line number will be added. Spaces are placed between the arguments as with the `:echomsg` command. When used inside a try conditional, the message is raised as an error exception instead (see |try-echoerr|). Example: > :echoerr "This script just failed!" < If you just want a highlighted message use `:echohl`. And to get a beep: > :exe "normal \" :echoc[onsole] {expr1} .. *:echoc* *:echoconsole* Intended for testing: works like `:echomsg` but when running in the GUI and started from a terminal write the text to stdout. *:eval* :eval {expr} Evaluate {expr} and discard the result. Example: > :eval Getlist()->Filter()->append('$') < The expression is supposed to have a side effect, since the resulting value is not used. In the example the `append()` call appends the List with text to the buffer. This is similar to `:call` but works with any expression. In |Vim9| script an expression without an effect will result in error *E1207* . This should help noticing mistakes. The command can be shortened to `:ev` or `:eva`, but these are hard to recognize and therefore not to be used. The command cannot be followed by "|" and another command, since "|" is seen as part of the expression. *:exe* *:execute* :exe[cute] {expr1} .. Executes the string that results from the evaluation of {expr1} as an Ex command. Multiple arguments are concatenated, with a space in between. To avoid the extra space use the ".." operator to concatenate strings into one argument. {expr1} is used as the processed command, command line editing keys are not recognized. Cannot be followed by a comment. Examples: > :execute "buffer" nextbuf :execute "normal" count .. "w" < ":execute" can be used to append a command to commands that don't accept a '|'. Example: > :execute '!ls' | echo "theend" < ":execute" is also a nice way to avoid having to type control characters in a Vim script for a ":normal" command: > :execute "normal ixxx\" < This has an character, see |expr-string|. Be careful to correctly escape special characters in file names. The |fnameescape()| function can be used for Vim commands, |shellescape()| for |:!| commands. Examples: > :execute "e " .. fnameescape(filename) :execute "!ls " .. shellescape(filename, 1) < Note: The executed string may be any command-line, but starting or ending "if", "while" and "for" does not always work, because when commands are skipped the ":execute" is not evaluated and Vim loses track of where blocks start and end. Also "break" and "continue" should not be inside ":execute". This example does not work, because the ":execute" is not evaluated and Vim does not see the "while", and gives an error for finding an ":endwhile": > :if 0 : execute 'while i > 5' : echo "test" : endwhile :endif < It is allowed to have a "while" or "if" command completely in the executed string: > :execute 'while i < 5 | echo i | let i = i + 1 | endwhile' < *:exe-comment* ":execute", ":echo" and ":echon" cannot be followed by a comment directly, because they see the '"' as the start of a string. But, you can use '|' followed by a comment. Example: > :echo "foo" | "this is a comment ============================================================================== 8. Exception handling *exception-handling* The Vim script language comprises an exception handling feature. This section explains how it can be used in a Vim script. Exceptions may be raised by Vim on an error or on interrupt, see |catch-errors| and |catch-interrupt|. You can also explicitly throw an exception by using the ":throw" command, see |throw-catch|. TRY CONDITIONALS *try-conditionals* Exceptions can be caught or can cause cleanup code to be executed. You can use a try conditional to specify catch clauses (that catch exceptions) and/or a finally clause (to be executed for cleanup). A try conditional begins with a |:try| command and ends at the matching |:endtry| command. In between, you can use a |:catch| command to start a catch clause, or a |:finally| command to start a finally clause. There may be none or multiple catch clauses, but there is at most one finally clause, which must not be followed by any catch clauses. The lines before the catch clauses and the finally clause is called a try block. > :try : ... : ... TRY BLOCK : ... :catch /{pattern}/ : ... : ... CATCH CLAUSE : ... :catch /{pattern}/ : ... : ... CATCH CLAUSE : ... :finally : ... : ... FINALLY CLAUSE : ... :endtry The try conditional allows to watch code for exceptions and to take the appropriate actions. Exceptions from the try block may be caught. Exceptions from the try block and also the catch clauses may cause cleanup actions. When no exception is thrown during execution of the try block, the control is transferred to the finally clause, if present. After its execution, the script continues with the line following the ":endtry". When an exception occurs during execution of the try block, the remaining lines in the try block are skipped. The exception is matched against the patterns specified as arguments to the ":catch" commands. The catch clause after the first matching ":catch" is taken, other catch clauses are not executed. The catch clause ends when the next ":catch", ":finally", or ":endtry" command is reached - whatever is first. Then, the finally clause (if present) is executed. When the ":endtry" is reached, the script execution continues in the following line as usual. When an exception that does not match any of the patterns specified by the ":catch" commands is thrown in the try block, the exception is not caught by that try conditional and none of the catch clauses is executed. Only the finally clause, if present, is taken. The exception pends during execution of the finally clause. It is resumed at the ":endtry", so that commands after the ":endtry" are not executed and the exception might be caught elsewhere, see |try-nesting|. When during execution of a catch clause another exception is thrown, the remaining lines in that catch clause are not executed. The new exception is not matched against the patterns in any of the ":catch" commands of the same try conditional and none of its catch clauses is taken. If there is, however, a finally clause, it is executed, and the exception pends during its execution. The commands following the ":endtry" are not executed. The new exception might, however, be caught elsewhere, see |try-nesting|. When during execution of the finally clause (if present) an exception is thrown, the remaining lines in the finally clause are skipped. If the finally clause has been taken because of an exception from the try block or one of the catch clauses, the original (pending) exception is discarded. The commands following the ":endtry" are not executed, and the exception from the finally clause is propagated and can be caught elsewhere, see |try-nesting|. The finally clause is also executed, when a ":break" or ":continue" for a ":while" loop enclosing the complete try conditional is executed from the try block or a catch clause. Or when a ":return" or ":finish" is executed from the try block or a catch clause of a try conditional in a function or sourced script, respectively. The ":break", ":continue", ":return", or ":finish" pends during execution of the finally clause and is resumed when the ":endtry" is reached. It is, however, discarded when an exception is thrown from the finally clause. When a ":break" or ":continue" for a ":while" loop enclosing the complete try conditional or when a ":return" or ":finish" is encountered in the finally clause, the rest of the finally clause is skipped, and the ":break", ":continue", ":return" or ":finish" is executed as usual. If the finally clause has been taken because of an exception or an earlier ":break", ":continue", ":return", or ":finish" from the try block or a catch clause, this pending exception or command is discarded. For examples see |throw-catch| and |try-finally|. NESTING OF TRY CONDITIONALS *try-nesting* Try conditionals can be nested arbitrarily. That is, a complete try conditional can be put into the try block, a catch clause, or the finally clause of another try conditional. If the inner try conditional does not catch an exception thrown in its try block or throws a new exception from one of its catch clauses or its finally clause, the outer try conditional is checked according to the rules above. If the inner try conditional is in the try block of the outer try conditional, its catch clauses are checked, but otherwise only the finally clause is executed. It does not matter for nesting, whether the inner try conditional is directly contained in the outer one, or whether the outer one sources a script or calls a function containing the inner try conditional. When none of the active try conditionals catches an exception, just their finally clauses are executed. Thereafter, the script processing terminates. An error message is displayed in case of an uncaught exception explicitly thrown by a ":throw" command. For uncaught error and interrupt exceptions implicitly raised by Vim, the error message(s) or interrupt message are shown as usual. For examples see |throw-catch|. EXAMINING EXCEPTION HANDLING CODE *except-examine* Exception handling code can get tricky. If you are in doubt what happens, set 'verbose' to 13 or use the ":13verbose" command modifier when sourcing your script file. Then you see when an exception is thrown, discarded, caught, or finished. When using a verbosity level of at least 14, things pending in a finally clause are also shown. This information is also given in debug mode (see |debug-scripts|). THROWING AND CATCHING EXCEPTIONS *throw-catch* You can throw any number or string as an exception. Use the |:throw| command and pass the value to be thrown as argument: > :throw 4711 :throw "string" < *throw-expression* You can also specify an expression argument. The expression is then evaluated first, and the result is thrown: > :throw 4705 + strlen("string") :throw strpart("strings", 0, 6) An exception might be thrown during evaluation of the argument of the ":throw" command. Unless it is caught there, the expression evaluation is abandoned. The ":throw" command then does not throw a new exception. Example: > :function! Foo(arg) : try : throw a:arg : catch /foo/ : endtry : return 1 :endfunction : :function! Bar() : echo "in Bar" : return 4710 :endfunction : :throw Foo("arrgh") + Bar() This throws "arrgh", and "in Bar" is not displayed since Bar() is not executed. > :throw Foo("foo") + Bar() however displays "in Bar" and throws 4711. Any other command that takes an expression as argument might also be abandoned by an (uncaught) exception during the expression evaluation. The exception is then propagated to the caller of the command. Example: > :if Foo("arrgh") : echo "then" :else : echo "else" :endif Here neither of "then" or "else" is displayed. *catch-order* Exceptions can be caught by a try conditional with one or more |:catch| commands, see |try-conditionals|. The values to be caught by each ":catch" command can be specified as a pattern argument. The subsequent catch clause gets executed when a matching exception is caught. Example: > :function! Foo(value) : try : throw a:value : catch /^\d\+$/ : echo "Number thrown" : catch /.*/ : echo "String thrown" : endtry :endfunction : :call Foo(0x1267) :call Foo('string') The first call to Foo() displays "Number thrown", the second "String thrown". An exception is matched against the ":catch" commands in the order they are specified. Only the first match counts. So you should place the more specific ":catch" first. The following order does not make sense: > : catch /.*/ : echo "String thrown" : catch /^\d\+$/ : echo "Number thrown" The first ":catch" here matches always, so that the second catch clause is never taken. *throw-variables* If you catch an exception by a general pattern, you may access the exact value in the variable |v:exception|: > : catch /^\d\+$/ : echo "Number thrown. Value is" v:exception You may also be interested where an exception was thrown. This is stored in |v:throwpoint|. Note that "v:exception" and "v:throwpoint" are valid for the exception most recently caught as long it is not finished. Example: > :function! Caught() : if v:exception != "" : echo 'Caught "' . v:exception .. '" in ' .. v:throwpoint : else : echo 'Nothing caught' : endif :endfunction : :function! Foo() : try : try : try : throw 4711 : finally : call Caught() : endtry : catch /.*/ : call Caught() : throw "oops" : endtry : catch /.*/ : call Caught() : finally : call Caught() : endtry :endfunction : :call Foo() This displays > Nothing caught Caught "4711" in function Foo, line 4 Caught "oops" in function Foo, line 10 Nothing caught A practical example: The following command ":LineNumber" displays the line number in the script or function where it has been used: > :function! LineNumber() : return substitute(v:throwpoint, '.*\D\(\d\+\).*', '\1', "") :endfunction :command! LineNumber try | throw "" | catch | echo LineNumber() | endtry < *try-nested* An exception that is not caught by a try conditional can be caught by a surrounding try conditional: > :try : try : throw "foo" : catch /foobar/ : echo "foobar" : finally : echo "inner finally" : endtry :catch /foo/ : echo "foo" :endtry The inner try conditional does not catch the exception, just its finally clause is executed. The exception is then caught by the outer try conditional. The example displays "inner finally" and then "foo". *throw-from-catch* You can catch an exception and throw a new one to be caught elsewhere from the catch clause: > :function! Foo() : throw "foo" :endfunction : :function! Bar() : try : call Foo() : catch /foo/ : echo "Caught foo, throw bar" : throw "bar" : endtry :endfunction : :try : call Bar() :catch /.*/ : echo "Caught" v:exception :endtry This displays "Caught foo, throw bar" and then "Caught bar". *rethrow* There is no real rethrow in the Vim script language, but you may throw "v:exception" instead: > :function! Bar() : try : call Foo() : catch /.*/ : echo "Rethrow" v:exception : throw v:exception : endtry :endfunction < *try-echoerr* Note that this method cannot be used to "rethrow" Vim error or interrupt exceptions, because it is not possible to fake Vim internal exceptions. Trying so causes an error exception. You should throw your own exception denoting the situation. If you want to cause a Vim error exception containing the original error exception value, you can use the |:echoerr| command: > :try : try : asdf : catch /.*/ : echoerr v:exception : endtry :catch /.*/ : echo v:exception :endtry This code displays Vim(echoerr):Vim:E492: Not an editor command: asdf ~ CLEANUP CODE *try-finally* Scripts often change global settings and restore them at their end. If the user however interrupts the script by pressing CTRL-C, the settings remain in an inconsistent state. The same may happen to you in the development phase of a script when an error occurs or you explicitly throw an exception without catching it. You can solve these problems by using a try conditional with a finally clause for restoring the settings. Its execution is guaranteed on normal control flow, on error, on an explicit ":throw", and on interrupt. (Note that errors and interrupts from inside the try conditional are converted to exceptions. When not caught, they terminate the script after the finally clause has been executed.) Example: > :try : let s:saved_ts = &ts : set ts=17 : : " Do the hard work here. : :finally : let &ts = s:saved_ts : unlet s:saved_ts :endtry This method should be used locally whenever a function or part of a script changes global settings which need to be restored on failure or normal exit of that function or script part. *break-finally* Cleanup code works also when the try block or a catch clause is left by a ":continue", ":break", ":return", or ":finish". Example: > :let first = 1 :while 1 : try : if first : echo "first" : let first = 0 : continue : else : throw "second" : endif : catch /.*/ : echo v:exception : break : finally : echo "cleanup" : endtry : echo "still in while" :endwhile :echo "end" This displays "first", "cleanup", "second", "cleanup", and "end". > :function! Foo() : try : return 4711 : finally : echo "cleanup\n" : endtry : echo "Foo still active" :endfunction : :echo Foo() "returned by Foo" This displays "cleanup" and "4711 returned by Foo". You don't need to add an extra ":return" in the finally clause. (Above all, this would override the return value.) *except-from-finally* Using either of ":continue", ":break", ":return", ":finish", or ":throw" in a finally clause is possible, but not recommended since it abandons the cleanup actions for the try conditional. But, of course, interrupt and error exceptions might get raised from a finally clause. Example where an error in the finally clause stops an interrupt from working correctly: > :try : try : echo "Press CTRL-C for interrupt" : while 1 : endwhile : finally : unlet novar : endtry :catch /novar/ :endtry :echo "Script still running" :sleep 1 If you need to put commands that could fail into a finally clause, you should think about catching or ignoring the errors in these commands, see |catch-errors| and |ignore-errors|. CATCHING ERRORS *catch-errors* If you want to catch specific errors, you just have to put the code to be watched in a try block and add a catch clause for the error message. The presence of the try conditional causes all errors to be converted to an exception. No message is displayed and |v:errmsg| is not set then. To find the right pattern for the ":catch" command, you have to know how the format of the error exception is. Error exceptions have the following format: > Vim({cmdname}):{errmsg} or > Vim:{errmsg} {cmdname} is the name of the command that failed; the second form is used when the command name is not known. {errmsg} is the error message usually produced when the error occurs outside try conditionals. It always begins with a capital "E", followed by a two or three-digit error number, a colon, and a space. Examples: The command > :unlet novar normally produces the error message > E108: No such variable: "novar" which is converted inside try conditionals to an exception > Vim(unlet):E108: No such variable: "novar" The command > :dwim normally produces the error message > E492: Not an editor command: dwim which is converted inside try conditionals to an exception > Vim:E492: Not an editor command: dwim You can catch all ":unlet" errors by a > :catch /^Vim(unlet):/ or all errors for misspelled command names by a > :catch /^Vim:E492:/ Some error messages may be produced by different commands: > :function nofunc and > :delfunction nofunc both produce the error message > E128: Function name must start with a capital: nofunc which is converted inside try conditionals to an exception > Vim(function):E128: Function name must start with a capital: nofunc or > Vim(delfunction):E128: Function name must start with a capital: nofunc respectively. You can catch the error by its number independently on the command that caused it if you use the following pattern: > :catch /^Vim(\a\+):E128:/ Some commands like > :let x = novar produce multiple error messages, here: > E121: Undefined variable: novar E15: Invalid expression: novar Only the first is used for the exception value, since it is the most specific one (see |except-several-errors|). So you can catch it by > :catch /^Vim(\a\+):E121:/ You can catch all errors related to the name "nofunc" by > :catch /\/ You can catch all Vim errors in the ":write" and ":read" commands by > :catch /^Vim(\(write\|read\)):E\d\+:/ You can catch all Vim errors by the pattern > :catch /^Vim\((\a\+)\)\=:E\d\+:/ < *catch-text* NOTE: You should never catch the error message text itself: > :catch /No such variable/ only works in the English locale, but not when the user has selected a different language by the |:language| command. It is however helpful to cite the message text in a comment: > :catch /^Vim(\a\+):E108:/ " No such variable IGNORING ERRORS *ignore-errors* You can ignore errors in a specific Vim command by catching them locally: > :try : write :catch :endtry But you are strongly recommended NOT to use this simple form, since it could catch more than you want. With the ":write" command, some autocommands could be executed and cause errors not related to writing, for instance: > :au BufWritePre * unlet novar There could even be such errors you are not responsible for as a script writer: a user of your script might have defined such autocommands. You would then hide the error from the user. It is much better to use > :try : write :catch /^Vim(write):/ :endtry which only catches real write errors. So catch only what you'd like to ignore intentionally. For a single command that does not cause execution of autocommands, you could even suppress the conversion of errors to exceptions by the ":silent!" command: > :silent! nunmap k This works also when a try conditional is active. CATCHING INTERRUPTS *catch-interrupt* When there are active try conditionals, an interrupt (CTRL-C) is converted to the exception "Vim:Interrupt". You can catch it like every exception. The script is not terminated, then. Example: > :function! TASK1() : sleep 10 :endfunction :function! TASK2() : sleep 20 :endfunction :while 1 : let command = input("Type a command: ") : try : if command == "" : continue : elseif command == "END" : break : elseif command == "TASK1" : call TASK1() : elseif command == "TASK2" : call TASK2() : else : echo "\nIllegal command:" command : continue : endif : catch /^Vim:Interrupt$/ : echo "\nCommand interrupted" : " Caught the interrupt. Continue with next prompt. : endtry :endwhile You can interrupt a task here by pressing CTRL-C; the script then asks for a new command. If you press CTRL-C at the prompt, the script is terminated. For testing what happens when CTRL-C would be pressed on a specific line in your script, use the debug mode and execute the |>quit| or |>interrupt| command on that line. See |debug-scripts|. CATCHING ALL *catch-all* The commands > :catch /.*/ :catch // :catch catch everything, error exceptions, interrupt exceptions and exceptions explicitly thrown by the |:throw| command. This is useful at the top level of a script in order to catch unexpected things. Example: > :try : : " do the hard work here : :catch /MyException/ : : " handle known problem : :catch /^Vim:Interrupt$/ : echo "Script interrupted" :catch /.*/ : echo "Internal error (" .. v:exception .. ")" : echo " - occurred at " .. v:throwpoint :endtry :" end of script Note: Catching all might catch more things than you want. Thus, you are strongly encouraged to catch only for problems that you can really handle by specifying a pattern argument to the ":catch". Example: Catching all could make it nearly impossible to interrupt a script by pressing CTRL-C: > :while 1 : try : sleep 1 : catch : endtry :endwhile EXCEPTIONS AND AUTOCOMMANDS *except-autocmd* Exceptions may be used during execution of autocommands. Example: > :autocmd User x try :autocmd User x throw "Oops!" :autocmd User x catch :autocmd User x echo v:exception :autocmd User x endtry :autocmd User x throw "Arrgh!" :autocmd User x echo "Should not be displayed" : :try : doautocmd User x :catch : echo v:exception :endtry This displays "Oops!" and "Arrgh!". *except-autocmd-Pre* For some commands, autocommands get executed before the main action of the command takes place. If an exception is thrown and not caught in the sequence of autocommands, the sequence and the command that caused its execution are abandoned and the exception is propagated to the caller of the command. Example: > :autocmd BufWritePre * throw "FAIL" :autocmd BufWritePre * echo "Should not be displayed" : :try : write :catch : echo "Caught:" v:exception "from" v:throwpoint :endtry Here, the ":write" command does not write the file currently being edited (as you can see by checking 'modified'), since the exception from the BufWritePre autocommand abandons the ":write". The exception is then caught and the script displays: > Caught: FAIL from BufWrite Auto commands for "*" < *except-autocmd-Post* For some commands, autocommands get executed after the main action of the command has taken place. If this main action fails and the command is inside an active try conditional, the autocommands are skipped and an error exception is thrown that can be caught by the caller of the command. Example: > :autocmd BufWritePost * echo "File successfully written!" : :try : write /i/m/p/o/s/s/i/b/l/e :catch : echo v:exception :endtry This just displays: > Vim(write):E212: Can't open file for writing (/i/m/p/o/s/s/i/b/l/e) If you really need to execute the autocommands even when the main action fails, trigger the event from the catch clause. Example: > :autocmd BufWritePre * set noreadonly :autocmd BufWritePost * set readonly : :try : write /i/m/p/o/s/s/i/b/l/e :catch : doautocmd BufWritePost /i/m/p/o/s/s/i/b/l/e :endtry < You can also use ":silent!": > :let x = "ok" :let v:errmsg = "" :autocmd BufWritePost * if v:errmsg != "" :autocmd BufWritePost * let x = "after fail" :autocmd BufWritePost * endif :try : silent! write /i/m/p/o/s/s/i/b/l/e :catch :endtry :echo x This displays "after fail". If the main action of the command does not fail, exceptions from the autocommands will be catchable by the caller of the command: > :autocmd BufWritePost * throw ":-(" :autocmd BufWritePost * echo "Should not be displayed" : :try : write :catch : echo v:exception :endtry < *except-autocmd-Cmd* For some commands, the normal action can be replaced by a sequence of autocommands. Exceptions from that sequence will be catchable by the caller of the command. Example: For the ":write" command, the caller cannot know whether the file had actually been written when the exception occurred. You need to tell it in some way. > :if !exists("cnt") : let cnt = 0 : : autocmd BufWriteCmd * if &modified : autocmd BufWriteCmd * let cnt = cnt + 1 : autocmd BufWriteCmd * if cnt % 3 == 2 : autocmd BufWriteCmd * throw "BufWriteCmdError" : autocmd BufWriteCmd * endif : autocmd BufWriteCmd * write | set nomodified : autocmd BufWriteCmd * if cnt % 3 == 0 : autocmd BufWriteCmd * throw "BufWriteCmdError" : autocmd BufWriteCmd * endif : autocmd BufWriteCmd * echo "File successfully written!" : autocmd BufWriteCmd * endif :endif : :try : write :catch /^BufWriteCmdError$/ : if &modified : echo "Error on writing (file contents not changed)" : else : echo "Error after writing" : endif :catch /^Vim(write):/ : echo "Error on writing" :endtry When this script is sourced several times after making changes, it displays first > File successfully written! then > Error on writing (file contents not changed) then > Error after writing etc. *except-autocmd-ill* You cannot spread a try conditional over autocommands for different events. The following code is ill-formed: > :autocmd BufWritePre * try : :autocmd BufWritePost * catch :autocmd BufWritePost * echo v:exception :autocmd BufWritePost * endtry : :write EXCEPTION HIERARCHIES AND PARAMETERIZED EXCEPTIONS *except-hier-param* Some programming languages allow to use hierarchies of exception classes or to pass additional information with the object of an exception class. You can do similar things in Vim. In order to throw an exception from a hierarchy, just throw the complete class name with the components separated by a colon, for instance throw the string "EXCEPT:MATHERR:OVERFLOW" for an overflow in a mathematical library. When you want to pass additional information with your exception class, add it in parentheses, for instance throw the string "EXCEPT:IO:WRITEERR(myfile)" for an error when writing "myfile". With the appropriate patterns in the ":catch" command, you can catch for base classes or derived classes of your hierarchy. Additional information in parentheses can be cut out from |v:exception| with the ":substitute" command. Example: > :function! CheckRange(a, func) : if a:a < 0 : throw "EXCEPT:MATHERR:RANGE(" .. a:func .. ")" : endif :endfunction : :function! Add(a, b) : call CheckRange(a:a, "Add") : call CheckRange(a:b, "Add") : let c = a:a + a:b : if c < 0 : throw "EXCEPT:MATHERR:OVERFLOW" : endif : return c :endfunction : :function! Div(a, b) : call CheckRange(a:a, "Div") : call CheckRange(a:b, "Div") : if (a:b == 0) : throw "EXCEPT:MATHERR:ZERODIV" : endif : return a:a / a:b :endfunction : :function! Write(file) : try : execute "write" fnameescape(a:file) : catch /^Vim(write):/ : throw "EXCEPT:IO(" .. getcwd() .. ", " .. a:file .. "):WRITEERR" : endtry :endfunction : :try : : " something with arithmetic and I/O : :catch /^EXCEPT:MATHERR:RANGE/ : let function = substitute(v:exception, '.*(\(\a\+\)).*', '\1', "") : echo "Range error in" function : :catch /^EXCEPT:MATHERR/ " catches OVERFLOW and ZERODIV : echo "Math error" : :catch /^EXCEPT:IO/ : let dir = substitute(v:exception, '.*(\(.\+\),\s*.\+).*', '\1', "") : let file = substitute(v:exception, '.*(.\+,\s*\(.\+\)).*', '\1', "") : if file !~ '^/' : let file = dir .. "/" .. file : endif : echo 'I/O error for "' .. file .. '"' : :catch /^EXCEPT/ : echo "Unspecified error" : :endtry The exceptions raised by Vim itself (on error or when pressing CTRL-C) use a flat hierarchy: they are all in the "Vim" class. You cannot throw yourself exceptions with the "Vim" prefix; they are reserved for Vim. Vim error exceptions are parameterized with the name of the command that failed, if known. See |catch-errors|. PECULIARITIES *except-compat* The exception handling concept requires that the command sequence causing the exception is aborted immediately and control is transferred to finally clauses and/or a catch clause. In the Vim script language there are cases where scripts and functions continue after an error: in functions without the "abort" flag or in a command after ":silent!", control flow goes to the following line, and outside functions, control flow goes to the line following the outermost ":endwhile" or ":endif". On the other hand, errors should be catchable as exceptions (thus, requiring the immediate abortion). This problem has been solved by converting errors to exceptions and using immediate abortion (if not suppressed by ":silent!") only when a try conditional is active. This is no restriction since an (error) exception can be caught only from an active try conditional. If you want an immediate termination without catching the error, just use a try conditional without catch clause. (You can cause cleanup code being executed before termination by specifying a finally clause.) When no try conditional is active, the usual abortion and continuation behavior is used instead of immediate abortion. This ensures compatibility of scripts written for Vim 6.1 and earlier. However, when sourcing an existing script that does not use exception handling commands (or when calling one of its functions) from inside an active try conditional of a new script, you might change the control flow of the existing script on error. You get the immediate abortion on error and can catch the error in the new script. If however the sourced script suppresses error messages by using the ":silent!" command (checking for errors by testing |v:errmsg| if appropriate), its execution path is not changed. The error is not converted to an exception. (See |:silent|.) So the only remaining cause where this happens is for scripts that don't care about errors and produce error messages. You probably won't want to use such code from your new scripts. *except-syntax-err* Syntax errors in the exception handling commands are never caught by any of the ":catch" commands of the try conditional they belong to. Its finally clauses, however, is executed. Example: > :try : try : throw 4711 : catch /\(/ : echo "in catch with syntax error" : catch : echo "inner catch-all" : finally : echo "inner finally" : endtry :catch : echo 'outer catch-all caught "' .. v:exception .. '"' : finally : echo "outer finally" :endtry This displays: > inner finally outer catch-all caught "Vim(catch):E54: Unmatched \(" outer finally The original exception is discarded and an error exception is raised, instead. *except-single-line* The ":try", ":catch", ":finally", and ":endtry" commands can be put on a single line, but then syntax errors may make it difficult to recognize the "catch" line, thus you better avoid this. Example: > :try | unlet! foo # | catch | endtry raises an error exception for the trailing characters after the ":unlet!" argument, but does not see the ":catch" and ":endtry" commands, so that the error exception is discarded and the "E488: Trailing characters" message gets displayed. *except-several-errors* When several errors appear in a single command, the first error message is usually the most specific one and therefore converted to the error exception. Example: > echo novar causes > E121: Undefined variable: novar E15: Invalid expression: novar The value of the error exception inside try conditionals is: > Vim(echo):E121: Undefined variable: novar < *except-syntax-error* But when a syntax error is detected after a normal error in the same command, the syntax error is used for the exception being thrown. Example: > unlet novar # causes > E108: No such variable: "novar" E488: Trailing characters The value of the error exception inside try conditionals is: > Vim(unlet):E488: Trailing characters This is done because the syntax error might change the execution path in a way not intended by the user. Example: > try try | unlet novar # | catch | echo v:exception | endtry catch /.*/ echo "outer catch:" v:exception endtry This displays "outer catch: Vim(unlet):E488: Trailing characters", and then a "E600: Missing :endtry" error message is given, see |except-single-line|. ============================================================================== 9. Examples *eval-examples* Printing in Binary ~ > :" The function Nr2Bin() returns the binary string representation of a number. :func Nr2Bin(nr) : let n = a:nr : let r = "" : while n : let r = '01'[n % 2] .. r : let n = n / 2 : endwhile : return r :endfunc :" The function String2Bin() converts each character in a string to a :" binary string, separated with dashes. :func String2Bin(str) : let out = '' : for ix in range(strlen(a:str)) : let out = out .. '-' .. Nr2Bin(char2nr(a:str[ix])) : endfor : return out[1:] :endfunc Example of its use: > :echo Nr2Bin(32) result: "100000" > :echo String2Bin("32") result: "110011-110010" Sorting lines ~ This example sorts lines with a specific compare function. > :func SortBuffer() : let lines = getline(1, '$') : call sort(lines, function("Strcmp")) : call setline(1, lines) :endfunction As a one-liner: > :call setline(1, sort(getline(1, '$'), function("Strcmp"))) scanf() replacement ~ *sscanf* There is no sscanf() function in Vim. If you need to extract parts from a line, you can use matchstr() and substitute() to do it. This example shows how to get the file name, line number and column number out of a line like "foobar.txt, 123, 45". > :" Set up the match bit :let mx='\(\f\+\),\s*\(\d\+\),\s*\(\d\+\)' :"get the part matching the whole expression :let l = matchstr(line, mx) :"get each item out of the match :let file = substitute(l, mx, '\1', '') :let lnum = substitute(l, mx, '\2', '') :let col = substitute(l, mx, '\3', '') The input is in the variable "line", the results in the variables "file", "lnum" and "col". (idea from Michael Geddes) getting the scriptnames in a Dictionary ~ *scriptnames-dictionary* The `:scriptnames` command can be used to get a list of all script files that have been sourced. There is also the `getscriptinfo()` function, but the information returned is not exactly the same. In case you need to manipulate the list, this code can be used as a base: > # Create or update scripts dictionary, indexed by SNR, and return it. def Scripts(scripts: dict = {}): dict for info in getscriptinfo() if scripts->has_key(info.sid) continue endif scripts[info.sid] = info.name endfor return scripts enddef ============================================================================== 10. Vim script versions *vimscript-version* *vimscript-versions* *scriptversion* Over time many features have been added to Vim script. This includes Ex commands, functions, variable types, etc. Each individual feature can be checked with the |has()| and |exists()| functions. Sometimes old syntax of functionality gets in the way of making Vim better. When support is taken away this will break older Vim scripts. To make this explicit the |:scriptversion| command can be used. When a Vim script is not compatible with older versions of Vim this will give an explicit error, instead of failing in mysterious ways. When using a legacy function, defined with `:function`, in |Vim9| script then scriptversion 4 is used. *scriptversion-1* > :scriptversion 1 < This is the original Vim script, same as not using a |:scriptversion| command. Can be used to go back to old syntax for a range of lines. Test for support with: > has('vimscript-1') < *scriptversion-2* > :scriptversion 2 < String concatenation with "." is not supported, use ".." instead. This avoids the ambiguity using "." for Dict member access and floating point numbers. Now ".5" means the number 0.5. *scriptversion-3* > :scriptversion 3 < All |vim-variable|s must be prefixed by "v:". E.g. "version" doesn't work as |v:version| anymore, it can be used as a normal variable. Same for some obvious names as "count" and others. Test for support with: > has('vimscript-3') < *scriptversion-4* > :scriptversion 4 < Numbers with a leading zero are not recognized as octal. "0o" or "0O" is still recognized as octal. With the previous version you get: > echo 017 " displays 15 (octal) echo 0o17 " displays 15 (octal) echo 018 " displays 18 (decimal) < with script version 4: > echo 017 " displays 17 (decimal) echo 0o17 " displays 15 (octal) echo 018 " displays 18 (decimal) < Also, it is possible to use single quotes inside numbers to make them easier to read: > echo 1'000'000 < The quotes must be surrounded by digits. Test for support with: > has('vimscript-4') ============================================================================== 11. No +eval feature *no-eval-feature* When the |+eval| feature was disabled at compile time, none of the expression evaluation commands are available. To prevent this from causing Vim scripts to generate all kinds of errors, the ":if" and ":endif" commands are still recognized, though the argument of the ":if" and everything between the ":if" and the matching ":endif" is ignored. Nesting of ":if" blocks is allowed, but only if the commands are at the start of the line. The ":else" command is not recognized. Example of how to avoid executing commands when the |+eval| feature is missing: > :if 1 : echo "Expression evaluation is compiled in" :else : echo "You will _never_ see this message" :endif To execute a command only when the |+eval| feature is disabled can be done in two ways. The simplest is to exit the script (or Vim) prematurely: > if 1 echo "commands executed with +eval" finish endif args " command executed without +eval If you do not want to abort loading the script you can use a trick, as this example shows: > silent! while 0 set history=111 silent! endwhile When the |+eval| feature is available the command is skipped because of the "while 0". Without the |+eval| feature the "while 0" is an error, which is silently ignored, and the command is executed. ============================================================================== 12. The sandbox *eval-sandbox* *sandbox* The 'foldexpr', 'formatexpr', 'includeexpr', 'indentexpr', 'statusline' and 'foldtext' options may be evaluated in a sandbox. This means that you are protected from these expressions having nasty side effects. This gives some safety for when these options are set from a modeline. It is also used when the command from a tags file is executed and for CTRL-R = in the command line. The sandbox is also used for the |:sandbox| command. *E48* These items are not allowed in the sandbox: - changing the buffer text - defining or changing mapping, autocommands, user commands - setting certain options (see |option-summary|) - setting certain v: variables (see |v:var|) *E794* - executing a shell command - reading or writing a file - jumping to another buffer or editing a file - executing Python, Perl, etc. commands This is not guaranteed 100% secure, but it should block most attacks. *:san* *:sandbox* :san[dbox] {cmd} Execute {cmd} in the sandbox. Useful to evaluate an option that may have been set from a modeline, e.g. 'foldexpr'. *sandbox-option* A few options contain an expression. When this expression is evaluated it may have to be done in the sandbox to avoid a security risk. But the sandbox is restrictive, thus this only happens when the option was set from an insecure location. Insecure in this context are: - sourcing a .vimrc or .exrc in the current directory - while executing in the sandbox - value coming from a modeline - executing a function that was defined in the sandbox Note that when in the sandbox and saving an option value and restoring it, the option will still be marked as it was set in the sandbox. ============================================================================== 13. Textlock *textlock* In a few situations it is not allowed to change the text in the buffer, jump to another window and some other things that might confuse or break what Vim is currently doing. This mostly applies to things that happen when Vim is actually doing something else. For example, evaluating the 'balloonexpr' may happen any moment the mouse cursor is resting at some position. This is not allowed when the textlock is active: - changing the buffer text - jumping to another buffer or window - editing another file - closing a window or quitting Vim - etc. ============================================================================== 14. Vim script library *vim-script-library* Vim comes bundled with a Vim script library, that can be used by runtime, script authors. Currently, it only includes very few functions, but it may grow over time. The functions are available as |Vim9-script| as well as using legacy Vim script (to be used for non Vim 9.0 versions and Neovim). *dist#vim* *dist#vim9* The functions make use of the autoloaded prefix "dist#vim" (for legacy Vim script and Neovim) and "dist#vim9" for Vim9 script. The following functions are available: dist#vim#IsSafeExecutable(filetype, executable) ~ dist#vim9#IsSafeExecutable(filetype:string, executable:string): bool ~ This function takes a filetype and an executable and checks whether it is safe to execute the given executable. For security reasons users may not want to have Vim execute random executables or may have forbidden to do so for specific filetypes by setting the "_exec" variable (|plugin_exec|). It returns |true| or |false| to indicate whether the plugin should run the given executable. It takes the following arguments: argument type ~ filetype string executable string vim:tw=78:ts=8:noet:ft=help:norl: