summaryrefslogtreecommitdiffstats
path: root/docs
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2021-02-26 04:10:02 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2021-02-26 04:10:02 +0000
commit227038a454943a075c51b60988c53f77a605fa10 (patch)
tree15d438e5519a2ba9ee7ba8c6f864bbc597ea33b5 /docs
parentInitial commit. (diff)
downloadptpython-227038a454943a075c51b60988c53f77a605fa10.tar.xz
ptpython-227038a454943a075c51b60988c53f77a605fa10.zip
Adding upstream version 3.0.16.upstream/3.0.16
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'docs')
-rw-r--r--docs/concurrency-challenges.rst91
-rw-r--r--docs/images/example1.pngbin0 -> 28700 bytes
-rw-r--r--docs/images/file-completion.pngbin0 -> 37364 bytes
-rw-r--r--docs/images/ipython.pngbin0 -> 36967 bytes
-rw-r--r--docs/images/multiline.pngbin0 -> 33129 bytes
-rw-r--r--docs/images/ptpython-history-help.pngbin0 -> 103337 bytes
-rw-r--r--docs/images/ptpython-menu.pngbin0 -> 60225 bytes
-rw-r--r--docs/images/ptpython.pngbin0 -> 28700 bytes
-rw-r--r--docs/images/validation.pngbin0 -> 17124 bytes
-rw-r--r--docs/images/windows.pngbin0 -> 18368 bytes
10 files changed, 91 insertions, 0 deletions
diff --git a/docs/concurrency-challenges.rst b/docs/concurrency-challenges.rst
new file mode 100644
index 0000000..b56d969
--- /dev/null
+++ b/docs/concurrency-challenges.rst
@@ -0,0 +1,91 @@
+
+Concurrency-related challenges regarding embedding of ptpython in asyncio code
+==============================================================================
+
+Things we want to be possible
+-----------------------------
+
+- Embed blocking ptpython in non-asyncio code (the normal use case).
+- Embed blocking ptpython in asyncio code (the event loop will block).
+- Embed awaitable ptpython in asyncio code (the loop will continue).
+- React to resize events (SIGWINCH).
+- Support top-level await.
+- Be able to patch_stdout, so that logging messages from another thread will be
+ printed above the prompt.
+- It should be possible to handle `KeyboardInterrupt` during evaluation of an
+ expression.
+- The "eval" should happen in the same thread from where embed() was called.
+
+
+Limitations of asyncio/python
+-----------------------------
+
+- We can only listen to SIGWINCH signal (resize) events in the main thread.
+
+- Usage of Control-C for triggering a `KeyboardInterrupt` only works for code
+ running in the main thread. (And only if the terminal was not set in raw
+ input mode).
+
+- Spawning a new event loop from within a coroutine, that's being executed in
+ an existing event loop is not allowed in asyncio. We can however spawn any
+ event loop in a separate thread, and wait for that thread to finish.
+
+- For patch_stdout to work correctly, we have to know what prompt_toolkit
+ application is running on the terminal, then tell that application to print
+ the output and redraw itself.
+
+
+Additional challenges for IPython
+---------------------------------
+
+IPython supports integration of 3rd party event loops (for various GUI
+toolkits). These event loops are supposed to continue running while we are
+prompting for input. In an asyncio environment, it means that there are
+situations where we have to juggle three event loops:
+
+- The asyncio loop in which the code was embedded.
+- The asyncio loop from the prompt.
+- The 3rd party GUI loop.
+
+Approach taken in ptpython 3.0.11
+---------------------------------
+
+For ptpython, the most reliable solution is to to run the prompt_toolkit input
+prompt in a separate background thread. This way it can use its own asyncio
+event loop without ever having to interfere with whatever runs in the main
+thread.
+
+Then, depending on how we embed, we do the following:
+When a normal blocking embed is used:
+ * We start the UI thread for the input, and do a blocking wait on
+ `thread.join()` here.
+ * The "eval" happens in the main thread.
+ * The "print" happens also in the main thread. Unless a pager is shown,
+ which is also a prompt_toolkit application, then the pager itself is runs
+ also in another thread, similar to the way we do the input.
+
+When an awaitable embed is used, for embedding in a coroutine, but having the
+event loop continue:
+ * We run the input method from the blocking embed in an asyncio executor
+ and do an `await loop.run_in_excecutor(...)`.
+ * The "eval" happens again in the main thread.
+ * "print" is also similar, except that the pager code (if used) runs in an
+ executor too.
+
+This means that the prompt_toolkit application code will always run in a
+different thread. It means it won't be able to respond to SIGWINCH (window
+resize events), but prompt_toolkit's 3.0.11 has now terminal size polling which
+solves this.
+
+Control-C key presses won't interrupt the main thread while we wait for input,
+because the prompt_toolkit application turns the terminal in raw mode, while
+it's reading, which means that it will receive control-c key presses as raw
+data in its own thread.
+
+Top-level await works in most situations as expected.
+- If a blocking embed is used. We execute ``loop.run_until_complete(code)``.
+ This assumes that the blocking embed is not used in a coroutine of a running
+ event loop, otherwise, this will attempt to start a nested event loop, which
+ asyncio does not support. In that case we will get an exception.
+- If an awaitable embed is used. We literally execute ``await code``. This will
+ integrate nicely in the current event loop.
diff --git a/docs/images/example1.png b/docs/images/example1.png
new file mode 100644
index 0000000..c46b55a
--- /dev/null
+++ b/docs/images/example1.png
Binary files differ
diff --git a/docs/images/file-completion.png b/docs/images/file-completion.png
new file mode 100644
index 0000000..3143293
--- /dev/null
+++ b/docs/images/file-completion.png
Binary files differ
diff --git a/docs/images/ipython.png b/docs/images/ipython.png
new file mode 100644
index 0000000..4ee77de
--- /dev/null
+++ b/docs/images/ipython.png
Binary files differ
diff --git a/docs/images/multiline.png b/docs/images/multiline.png
new file mode 100644
index 0000000..cdfeb32
--- /dev/null
+++ b/docs/images/multiline.png
Binary files differ
diff --git a/docs/images/ptpython-history-help.png b/docs/images/ptpython-history-help.png
new file mode 100644
index 0000000..a52b5c2
--- /dev/null
+++ b/docs/images/ptpython-history-help.png
Binary files differ
diff --git a/docs/images/ptpython-menu.png b/docs/images/ptpython-menu.png
new file mode 100644
index 0000000..923ca12
--- /dev/null
+++ b/docs/images/ptpython-menu.png
Binary files differ
diff --git a/docs/images/ptpython.png b/docs/images/ptpython.png
new file mode 100644
index 0000000..c46b55a
--- /dev/null
+++ b/docs/images/ptpython.png
Binary files differ
diff --git a/docs/images/validation.png b/docs/images/validation.png
new file mode 100644
index 0000000..cce14a1
--- /dev/null
+++ b/docs/images/validation.png
Binary files differ
diff --git a/docs/images/windows.png b/docs/images/windows.png
new file mode 100644
index 0000000..188d698
--- /dev/null
+++ b/docs/images/windows.png
Binary files differ