summaryrefslogtreecommitdiffstats
path: root/doc/clzip.texi
blob: 64e09e2abcd09700089d738fa32affdbe0d6ca76 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
\input texinfo @c -*-texinfo-*-
@c %**start of header
@setfilename clzip.info
@documentencoding ISO-8859-15
@settitle Clzip Manual
@finalout
@c %**end of header

@set UPDATED 6 February 2018
@set VERSION 1.10

@dircategory Data Compression
@direntry
* Clzip: (clzip).               LZMA lossless data compressor
@end direntry


@ifnothtml
@titlepage
@title Clzip
@subtitle LZMA lossless data compressor
@subtitle for Clzip version @value{VERSION}, @value{UPDATED}
@author by Antonio Diaz Diaz

@page
@vskip 0pt plus 1filll
@end titlepage

@contents
@end ifnothtml

@node Top
@top

This manual is for Clzip (version @value{VERSION}, @value{UPDATED}).

@menu
* Introduction::           Purpose and features of clzip
* Output::                 Meaning of clzip's output
* Invoking clzip::         Command line interface
* Quality assurance::      Design, development and testing of lzip
* File format::            Detailed format of the compressed file
* Algorithm::              How clzip compresses the data
* Stream format::          Format of the LZMA stream in lzip files
* Trailing data::          Extra data appended to the file
* Examples::               A small tutorial with examples
* Problems::               Reporting bugs
* Reference source code::  Source code illustrating stream format
* Concept index::          Index of concepts
@end menu

@sp 1
Copyright @copyright{} 2010-2018 Antonio Diaz Diaz.

This manual is free documentation: you have unlimited permission
to copy, distribute and modify it.


@node Introduction
@chapter Introduction
@cindex introduction

Clzip is a C language version of lzip, fully compatible with lzip-1.4 or
newer. As clzip is written in C, it may be easier to integrate in
applications like package managers, embedded devices, or systems lacking
a C++ compiler.

Lzip is a lossless data compressor with a user interface similar to the
one of gzip or bzip2. Lzip can compress about as fast as gzip
@w{(lzip -0)}, or compress most files more than bzip2 @w{(lzip -9)}.
Decompression speed is intermediate between gzip and bzip2. Lzip is
better than gzip and bzip2 from a data recovery perspective.

The lzip file format is designed for data sharing and long-term
archiving, taking into account both data integrity and decoder
availability:

@itemize @bullet
@item
The lzip format provides very safe integrity checking and some data
recovery means. The
@uref{http://www.nongnu.org/lzip/manual/lziprecover_manual.html#Data-safety,,lziprecover}
program can repair bit flip errors (one of the most common forms of data
corruption) in lzip files, and provides data recovery capabilities,
including error-checked merging of damaged copies of a file.
@ifnothtml
@xref{Data safety,,,lziprecover}.
@end ifnothtml

@item
The lzip format is as simple as possible (but not simpler). The lzip
manual provides the source code of a simple decompressor along with a
detailed explanation of how it works, so that with the only help of the
lzip manual it would be possible for a digital archaeologist to extract
the data from a lzip file long after quantum computers eventually render
LZMA obsolete.

@item
Additionally the lzip reference implementation is copylefted, which
guarantees that it will remain free forever.
@end itemize

A nice feature of the lzip format is that a corrupt byte is easier to
repair the nearer it is from the beginning of the file. Therefore, with
the help of lziprecover, losing an entire archive just because of a
corrupt byte near the beginning is a thing of the past.

The member trailer stores the 32-bit CRC of the original data, the size
of the original data and the size of the member. These values, together
with the end-of-stream marker, provide a 3 factor integrity checking
which guarantees that the decompressed version of the data is identical
to the original. This guards against corruption of the compressed data,
and against undetected bugs in clzip (hopefully very unlikely). The
chances of data corruption going undetected are microscopic. Be aware,
though, that the check occurs upon decompression, so it can only tell
you that something is wrong. It can't help you recover the original
uncompressed data.

Clzip uses the same well-defined exit status values used by lzip and
bzip2, which makes it safer than compressors returning ambiguous warning
values (like gzip) when it is used as a back end for other programs like
tar or zutils.

Clzip will automatically use the smallest possible dictionary size for
each file without exceeding the given limit. Keep in mind that the
decompression memory requirement is affected at compression time by the
choice of dictionary size limit.

The amount of memory required for compression is about 1 or 2 times the
dictionary size limit (1 if input file size is less than dictionary size
limit, else 2) plus 9 times the dictionary size really used. The option
@samp{-0} is special and only requires about @w{1.5 MiB} at most. The
amount of memory required for decompression is about @w{46 kB} larger
than the dictionary size really used.

When compressing, clzip replaces every file given in the command line
with a compressed version of itself, with the name "original_name.lz".
When decompressing, clzip attempts to guess the name for the decompressed
file from that of the compressed file as follows:

@multitable {anyothername} {becomes} {anyothername.out}
@item filename.lz  @tab becomes @tab filename
@item filename.tlz @tab becomes @tab filename.tar
@item anyothername @tab becomes @tab anyothername.out
@end multitable

(De)compressing a file is much like copying or moving it; therefore clzip
preserves the access and modification dates, permissions, and, when
possible, ownership of the file just as "cp -p" does. (If the user ID or
the group ID can't be duplicated, the file permission bits S_ISUID and
S_ISGID are cleared).

Clzip is able to read from some types of non regular files if the
@samp{--stdout} option is specified.

If no file names are specified, clzip compresses (or decompresses) from
standard input to standard output. In this case, clzip will decline to
write compressed output to a terminal, as this would be entirely
incomprehensible and therefore pointless.

Clzip will correctly decompress a file which is the concatenation of two
or more compressed files. The result is the concatenation of the
corresponding decompressed files. Integrity testing of concatenated
compressed files is also supported.

Clzip can produce multimember files, and lziprecover can safely recover
the undamaged members in case of file damage. Clzip can also split the
compressed output in volumes of a given size, even when reading from
standard input. This allows the direct creation of multivolume
compressed tar archives.

Clzip is able to compress and decompress streams of unlimited size by
automatically creating multimember output. The members so created are
large, about @w{2 PiB} each.


@node Output
@chapter Meaning of clzip's output
@cindex output

The output of clzip looks like this:

@example
clzip -v foo
  foo:  6.676:1, 14.98% ratio, 85.02% saved, 450560 in, 67493 out.

clzip -tvv foo.lz
  foo.lz:  6.676:1, 14.98% ratio, 85.02% saved.  ok
@end example

The meaning of each field is as follows:

@table @code
@item N:1
The compression ratio @w{(uncompressed_size / compressed_size)}, shown
as N to 1.

@item ratio
The inverse compression ratio @w{(compressed_size / uncompressed_size)},
shown as a percentage. A decimal ratio is easily obtained by moving the
decimal point two places to the left; @w{14.98% = 0.1498}.

@item saved
The space saved by compression @w{(1 - ratio)}, shown as a percentage.

@item in
The size of the uncompressed data. When decompressing or testing, it is
shown as @code{decompressed}. Note that clzip always prints the
uncompressed size before the compressed size when compressing,
decompressing, testing or listing.

@item out
The size of the compressed data. When decompressing or testing, it is
shown as @code{compressed}.

@end table

When decompressing or testing at verbosity level 4 (-vvvv), the
dictionary size used to compress the file and the CRC32 of the
uncompressed data are also shown.

LANGUAGE NOTE: Uncompressed = not compressed = plain data; it may never
have been compressed. Decompressed is used to refer to data which have
undergone the process of decompression.


@node Invoking clzip
@chapter Invoking clzip
@cindex invoking
@cindex options
@cindex usage
@cindex version

The format for running clzip is:

@example
clzip [@var{options}] [@var{files}]
@end example

@noindent
@samp{-} used as a @var{file} argument means standard input. It can be
mixed with other @var{files} and is read just once, the first time it
appears in the command line.

clzip supports the following options:

@table @code
@item -h
@itemx --help
Print an informative help message describing the options and exit.

@item -V
@itemx --version
Print the version number of clzip on the standard output and exit.

@anchor{--trailing-error}
@item -a
@itemx --trailing-error
Exit with error status 2 if any remaining input is detected after
decompressing the last member. Such remaining input is usually trailing
garbage that can be safely ignored. @xref{concat-example}.

@item -b @var{bytes}
@itemx --member-size=@var{bytes}
When compressing, set the member size limit to @var{bytes}. A small
member size may degrade compression ratio, so use it only when needed.
Valid values range from @w{100 kB} to @w{2 PiB}. Defaults to @w{2 PiB}.

@item -c
@itemx --stdout
Compress or decompress to standard output; keep input files unchanged.
If compressing several files, each file is compressed independently.
This option is needed when reading from a named pipe (fifo) or from a
device. Use it also to recover as much of the decompressed data as
possible when decompressing a corrupt file.

@item -d
@itemx --decompress
Decompress the specified files. If a file does not exist or can't be
opened, clzip continues decompressing the rest of the files. If a file
fails to decompress, or is a terminal, clzip exits immediately without
decompressing the rest of the files.

@item -f
@itemx --force
Force overwrite of output files.

@item -F
@itemx --recompress
When compressing, force re-compression of files whose name already has
the @samp{.lz} or @samp{.tlz} suffix.

@item -k
@itemx --keep
Keep (don't delete) input files during compression or decompression.

@item -l
@itemx --list
Print the uncompressed size, compressed size and percentage saved of the
specified files. Trailing data are ignored. The values produced are
correct even for multimember files. If more than one file is given, a
final line containing the cumulative sizes is printed. With @samp{-v},
the dictionary size, the number of members in the file, and the amount
of trailing data (if any) are also printed. With @samp{-vv}, the
positions and sizes of each member in multimember files are also
printed. @samp{-lq} can be used to verify quickly (without
decompressing) the structural integrity of the specified files. (Use
@samp{--test} to verify the data integrity). @samp{-alq} additionally
verifies that none of the specified files contain trailing data.

@item -m @var{bytes}
@itemx --match-length=@var{bytes}
When compressing, set the match length limit in bytes. After a match
this long is found, the search is finished. Valid values range from 5 to
273. Larger values usually give better compression ratios but longer
compression times.

@item -o @var{file}
@itemx --output=@var{file}
When reading from standard input and @samp{--stdout} has not been
specified, use @samp{@var{file}} as the virtual name of the uncompressed
file. This produces a file named @samp{@var{file}} when decompressing,
or a file named @samp{@var{file}.lz} when compressing. A second
@samp{.lz} extension is not added if @samp{@var{file}} already ends in
@samp{.lz} or @samp{.tlz}. When compressing and splitting the output in
volumes, several files named @samp{@var{file}00001.lz},
@samp{@var{file}00002.lz}, etc, are created.

@item -q
@itemx --quiet
Quiet operation. Suppress all messages.

@item -s @var{bytes}
@itemx --dictionary-size=@var{bytes}
When compressing, set the dictionary size limit in bytes. Clzip will use
the smallest possible dictionary size for each file without exceeding
this limit. Valid values range from @w{4 KiB} to @w{512 MiB}. Values 12
to 29 are interpreted as powers of two, meaning 2^12 to 2^29 bytes. Note
that dictionary sizes are quantized. If the specified size does not
match one of the valid sizes, it will be rounded upwards by adding up to
@w{(@var{bytes} / 8)} to it.

For maximum compression you should use a dictionary size limit as large
as possible, but keep in mind that the decompression memory requirement
is affected at compression time by the choice of dictionary size limit.

@item -S @var{bytes}
@itemx --volume-size=@var{bytes}
When compressing, split the compressed output into several volume files
with names @samp{original_name00001.lz}, @samp{original_name00002.lz},
etc, and set the volume size limit to @var{bytes}. Input files are kept
unchanged. Each volume is a complete, maybe multimember, lzip file. A
small volume size may degrade compression ratio, so use it only when
needed. Valid values range from @w{100 kB} to @w{4 EiB}.

@item -t
@itemx --test
Check integrity of the specified files, but don't decompress them. This
really performs a trial decompression and throws away the result. Use it
together with @samp{-v} to see information about the files. If a file
fails the test, does not exist, can't be opened, or is a terminal, clzip
continues checking the rest of the files. A final diagnostic is shown at
verbosity level 1 or higher if any file fails the test when testing
multiple files.

@item -v
@itemx --verbose
Verbose mode.@*
When compressing, show the compression ratio and size for each file
processed.@*
When decompressing or testing, further -v's (up to 4) increase the
verbosity level, showing status, compression ratio, dictionary size,
trailer contents (CRC, data size, member size), and up to 6 bytes of
trailing data (if any) both in hexadecimal and as a string of printable
ASCII characters.@*
Two or more @samp{-v} options show the progress of (de)compression.

@item -0 .. -9
Set the compression parameters (dictionary size and match length limit)
as shown in the table below. The default compression level is @samp{-6}.
Note that @samp{-9} can be much slower than @samp{-0}. These options
have no effect when decompressing, testing or listing.

The bidimensional parameter space of LZMA can't be mapped to a linear
scale optimal for all files. If your files are large, very repetitive,
etc, you may need to use the @samp{--dictionary-size} and
@samp{--match-length} options directly to achieve optimal performance.

@multitable {Level} {Dictionary size} {Match length limit}
@item Level @tab Dictionary size @tab Match length limit
@item -0 @tab 64 KiB @tab  16 bytes
@item -1 @tab  1 MiB @tab   5 bytes
@item -2 @tab  1.5 MiB @tab   6 bytes
@item -3 @tab  2 MiB @tab   8 bytes
@item -4 @tab  3 MiB @tab  12 bytes
@item -5 @tab  4 MiB @tab  20 bytes
@item -6 @tab  8 MiB @tab  36 bytes
@item -7 @tab 16 MiB @tab  68 bytes
@item -8 @tab 24 MiB @tab 132 bytes
@item -9 @tab 32 MiB @tab 273 bytes
@end multitable

@item --fast
@itemx --best
Aliases for GNU gzip compatibility.

@item --loose-trailing
When decompressing, testing or listing, allow trailing data whose first
bytes are so similar to the magic bytes of a lzip header that they can
be confused with a corrupt header. Use this option if a file triggers a
"corrupt header" error and the cause is not indeed a corrupt header.

@end table

Numbers given as arguments to options may be followed by a multiplier
and an optional @samp{B} for "byte".

Table of SI and binary prefixes (unit multipliers):

@multitable {Prefix} {kilobyte  (10^3 = 1000)} {|} {Prefix} {kibibyte (2^10 = 1024)}
@item Prefix @tab Value               @tab | @tab Prefix @tab Value
@item k @tab kilobyte  (10^3 = 1000)  @tab | @tab Ki @tab kibibyte (2^10 = 1024)
@item M @tab megabyte  (10^6)         @tab | @tab Mi @tab mebibyte (2^20)
@item G @tab gigabyte  (10^9)         @tab | @tab Gi @tab gibibyte (2^30)
@item T @tab terabyte  (10^12)        @tab | @tab Ti @tab tebibyte (2^40)
@item P @tab petabyte  (10^15)        @tab | @tab Pi @tab pebibyte (2^50)
@item E @tab exabyte   (10^18)        @tab | @tab Ei @tab exbibyte (2^60)
@item Z @tab zettabyte (10^21)        @tab | @tab Zi @tab zebibyte (2^70)
@item Y @tab yottabyte (10^24)        @tab | @tab Yi @tab yobibyte (2^80)
@end multitable

@sp 1
Exit status: 0 for a normal exit, 1 for environmental problems (file not
found, invalid flags, I/O errors, etc), 2 to indicate a corrupt or
invalid input file, 3 for an internal consistency error (eg, bug) which
caused clzip to panic.


@node Quality assurance
@chapter Design, development and testing of lzip
@cindex quality assurance

There are two ways of constructing a software design: One way is to make
it so simple that there are obviously no deficiencies and the other way
is to make it so complicated that there are no obvious deficiencies. The
first method is far more difficult.@*
--- C.A.R. Hoare

Lzip has been designed, written and tested with great care to be the
standard general-purpose compressor for unix-like systems. This chapter
describes the lessons learned from previous compressors (gzip and
bzip2), and their application to the design of lzip.

@sp 1
@section Format design

When gzip was designed in 1992, computers and operating systems were
much less capable than they are today. Gzip tried to work around some of
those limitations, like 8.3 file names, with additional fields in its
file format.

Today those limitations have mostly disappeared, and the format of gzip
has proved to be unnecessarily complicated. It includes fields that were
never used, others that have lost their usefulness, and finally others
that have become too limited.

Bzip2 was designed 5 years later, and its format is simpler than the one
of gzip.

Probably the worst defect of the gzip format from the point of view of
data safety is the variable size of its header. If the byte at offset 3
(flags) of a gzip member gets corrupted, it may become difficult to
recover the data, even if the compressed blocks are intact, because it
can't be known with certainty where the compressed blocks begin.

By contrast, the header of a lzip member has a fixed length of 6. The
LZMA stream in a lzip member always starts at offset 6, making it
trivial to recover the data even if the whole header becomes corrupt.

Bzip2 also provides a header of fixed length and marks the begin and end
of each compressed block with six magic bytes, making it possible to
find the compressed blocks even in case of file damage. But bzip2 does
not store the size of each compressed block, as lzip does.

Lzip provides better data recovery capabilities than any other gzip-like
compressor because its format has been designed from the beginning to be
simple and safe. It also helps that the LZMA data stream as used by lzip
is extraordinarily safe. It provides embedded error detection. Any
distance larger than the dictionary size acts as a forbidden symbol,
allowing the decompressor to detect the approximate position of errors,
and leaving very little work for the check sequence (CRC and data sizes)
in the detection of errors. Lzip is usually able to detect all posible
bit flips in the compressed data without resorting to the check
sequence. It would be difficult to write an automatic recovery tool like
lziprecover for the gzip format. And, as far as I know, it has never
been written.

Lzip, like gzip and bzip2, uses a CRC32 to check the integrity of the
decompressed data because it provides more accurate error detection than
CRC64 up to a compressed size of about @w{16 GiB}, a size larger than
that of most files. In the case of lzip, the additional detection
capability of the decompressor reduces the probability of undetected
errors more than a million times beyond what the CRC32 alone provides.

The lzip format is designed for long-term archiving. Therefore it
excludes any unneeded features that may interfere with the future
extraction of the decompressed data.

@sp 1
@subsection Gzip format (mis)features not present in lzip

@table @samp
@item Multiple algorithms

Gzip provides a CM (Compression Method) field that has never been used
because it is a bad idea to begin with. New compression methods may
require additional fields, making it impossible to implement new methods
and, at the same time, keep the same format. This field does not solve
the problem of format proliferation; it just makes the problem less
obvious.

@item Optional fields in header

Unless special precautions are taken, optional fields are generally a
bad idea because they produce a header of variable size. The gzip header
has 2 fields that, in addition to being optional, are zero-terminated.
This means that if any byte inside the field gets zeroed, or if the
terminating zero gets altered, gzip won't be able to find neither the
header CRC nor the compressed blocks.

@item Optional CRC for the header

Using an optional CRC for the header is not only a bad idea, it is an
error; it circumvents the HD of the CRC and may prevent the extraction
of perfectly good data. For example, if the CRC is used and the bit
enabling it is reset by a bit flip, the header will appear to be intact
(in spite of being corrupt) while the compressed blocks will appear to
be totally unrecoverable (in spite of being intact). Very misleading
indeed.

@item Metadata

The gzip format stores some metadata, like the modification time of the
original file or the operating system on which compression took place.
This complicates reproducible compression (obtaining identical
compressed output from identical input).

@end table

@subsection Lzip format improvements over gzip and bzip2

@table @samp
@item 64-bit size field

Probably the most frequently reported shortcoming of the gzip format is
that it only stores the least significant 32 bits of the uncompressed
size. The size of any file larger than @w{4 GiB} gets truncated.

Bzip2 does not store the uncompressed size of the file.

The lzip format provides a 64-bit field for the uncompressed size.
Additionaly, lzip produces multimember output automatically when the
size is too large for a single member, allowing for an unlimited
uncompressed size.

@item Distributed index

The lzip format provides a distributed index that, among other things,
helps plzip to decompress several times faster than pigz and helps
lziprecover do its job. Neither the gzip format nor the bzip2 format do
provide an index.

A distributed index is safer and more scalable than a monolithic index.
The monolithic index introduces a single point of failure in the
compressed file and may limit the number of members or the total
uncompressed size.

@end table

@section Quality of implementation

@table @samp
@item Accurate and robust error detection

The lzip format provides 3 factor integrity checking and the
decompressors report mismatches in each factor separately. This way if
just one byte in one factor fails but the other two factors match the
data, it probably means that the data are intact and the corruption just
affects the mismatching factor (CRC or data size) in the check sequence.

@item Multiple implementations

Just like the lzip format provides 3 factor protection against
undetected data corruption, the development methodology of the lzip
family of compressors provides 3 factor protection against undetected
programming errors.

Three related but independent compressor implementations, lzip, clzip
and minilzip/lzlib, are developed concurrently. Every stable release of
any of them is subjected to a hundred hours of intensive testing to
verify that it produces identical output to the other two. This
guarantees that all three implement the same algorithm, and makes it
unlikely that any of them may contain serious undiscovered errors. In
fact, no errors have been discovered in lzip since 2009.

Additionally, the three implementations have been extensively tested
with
@uref{http://www.nongnu.org/lzip/manual/lziprecover_manual.html#Unzcrash,,unzcrash},
valgrind and @samp{american fuzzy lop} without finding a single
vulnerability or false negative.
@ifnothtml
@xref{Unzcrash,,,lziprecover}.
@end ifnothtml

@item Dictionary size

Lzip automatically uses the smallest possible dictionary size for each
file. In addition to reducing the amount of memory required for
decompression, this feature also minimizes the probability of being
affected by RAM errors during compression.

@item Exit status

Returning a warning status of 2 is a design flaw of compress that leaked
into the design of gzip. Both bzip2 and lzip are free from this flaw.

@end table


@node File format
@chapter File format
@cindex file format

Perfection is reached, not when there is no longer anything to add, but
when there is no longer anything to take away.@*
--- Antoine de Saint-Exupery

@sp 1
In the diagram below, a box like this:
@verbatim
+---+
|   | <-- the vertical bars might be missing
+---+
@end verbatim

represents one byte; a box like this:
@verbatim
+==============+
|              |
+==============+
@end verbatim

represents a variable number of bytes.

@sp 1
A lzip file consists of a series of "members" (compressed data sets).
The members simply appear one after another in the file, with no
additional information before, between, or after them.

Each member has the following structure:
@verbatim
+--+--+--+--+----+----+=============+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| ID string | VN | DS | LZMA stream | CRC32 |   Data size   |  Member size  |
+--+--+--+--+----+----+=============+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
@end verbatim

All multibyte values are stored in little endian order.

@table @samp
@item ID string (the "magic" bytes)
A four byte string, identifying the lzip format, with the value "LZIP"
(0x4C, 0x5A, 0x49, 0x50).

@item VN (version number, 1 byte)
Just in case something needs to be modified in the future. 1 for now.

@item DS (coded dictionary size, 1 byte)
The dictionary size is calculated by taking a power of 2 (the base size)
and substracting from it a fraction between 0/16 and 7/16 of the base
size.@*
Bits 4-0 contain the base 2 logarithm of the base size (12 to 29).@*
Bits 7-5 contain the numerator of the fraction (0 to 7) to substract
from the base size to obtain the dictionary size.@*
Example: 0xD3 = 2^19 - 6 * 2^15 = 512 KiB - 6 * 32 KiB = 320 KiB@*
Valid values for dictionary size range from 4 KiB to 512 MiB.

@item LZMA stream
The LZMA stream, finished by an end of stream marker. Uses default
values for encoder properties. @xref{Stream format}, for a complete
description.

@item CRC32 (4 bytes)
CRC of the uncompressed original data.

@item Data size (8 bytes)
Size of the uncompressed original data.

@item Member size (8 bytes)
Total size of the member, including header and trailer. This field acts
as a distributed index, allows the verification of stream integrity, and
facilitates safe recovery of undamaged members from multimember files.

@end table


@node Algorithm
@chapter Algorithm
@cindex algorithm

In spite of its name (Lempel-Ziv-Markov chain-Algorithm), LZMA is not a
concrete algorithm; it is more like "any algorithm using the LZMA coding
scheme". For example, the option @samp{-0} of lzip uses the scheme in almost
the simplest way possible; issuing the longest match it can find, or a
literal byte if it can't find a match. Inversely, a much more elaborated
way of finding coding sequences of minimum size than the one currently
used by lzip could be developed, and the resulting sequence could also
be coded using the LZMA coding scheme.

Clzip currently implements two variants of the LZMA algorithm; fast
(used by option @samp{-0}) and normal (used by all other compression levels).

The high compression of LZMA comes from combining two basic, well-proven
compression ideas: sliding dictionaries (LZ77/78) and markov models (the
thing used by every compression algorithm that uses a range encoder or
similar order-0 entropy coder as its last stage) with segregation of
contexts according to what the bits are used for.

Clzip is a two stage compressor. The first stage is a Lempel-Ziv coder,
which reduces redundancy by translating chunks of data to their
corresponding distance-length pairs. The second stage is a range encoder
that uses a different probability model for each type of data;
distances, lengths, literal bytes, etc.

Here is how it works, step by step:

1) The member header is written to the output stream.

2) The first byte is coded literally, because there are no previous
bytes to which the match finder can refer to.

3) The main encoder advances to the next byte in the input data and
calls the match finder.

4) The match finder fills an array with the minimum distances before the
current byte where a match of a given length can be found.

5) Go back to step 3 until a sequence (formed of pairs, repeated
distances and literal bytes) of minimum price has been formed. Where the
price represents the number of output bits produced.

6) The range encoder encodes the sequence produced by the main encoder
and sends the produced bytes to the output stream.

7) Go back to step 3 until the input data are finished or until the
member or volume size limits are reached.

8) The range encoder is flushed.

9) The member trailer is written to the output stream.

10) If there are more data to compress, go back to step 1.

@sp 1
@noindent
The ideas embodied in clzip are due to (at least) the following people:
Abraham Lempel and Jacob Ziv (for the LZ algorithm), Andrey Markov (for
the definition of Markov chains), G.N.N. Martin (for the definition of
range encoding), Igor Pavlov (for putting all the above together in
LZMA), and Julian Seward (for bzip2's CLI).


@node Stream format
@chapter Format of the LZMA stream in lzip files
@cindex format of the LZMA stream

The LZMA algorithm has three parameters, called "special LZMA
properties", to adjust it for some kinds of binary data. These
parameters are; @samp{literal_context_bits} (with a default value of 3),
@samp{literal_pos_state_bits} (with a default value of 0), and
@samp{pos_state_bits} (with a default value of 2). As a general purpose
compressor, lzip only uses the default values for these parameters. In
particular @samp{literal_pos_state_bits} has been optimized away and
does not even appear in the code.

Lzip also finishes the LZMA stream with an "End Of Stream" marker (the
distance-length pair 0xFFFFFFFFU, 2), which in conjunction with the
"member size" field in the member trailer allows the verification of
stream integrity. The LZMA stream in lzip files always has these two
features (default properties and EOS marker) and is referred to in this
document as LZMA-302eos or LZMA-lzip.

The second stage of LZMA is a range encoder that uses a different
probability model for each type of symbol; distances, lengths, literal
bytes, etc. Range encoding conceptually encodes all the symbols of the
message into one number. Unlike Huffman coding, which assigns to each
symbol a bit-pattern and concatenates all the bit-patterns together,
range encoding can compress one symbol to less than one bit. Therefore
the compressed data produced by a range encoder can't be split in pieces
that could be individually described.

It seems that the only way of describing the LZMA-302eos stream is
describing the algorithm that decodes it. And given the many details
about the range decoder that need to be described accurately, the source
code of a real decoder seems the only appropriate reference to use.

What follows is a description of the decoding algorithm for LZMA-302eos
streams using as reference the source code of "lzd", an educational
decompressor for lzip files which can be downloaded from the lzip
download directory. The source code of lzd is included in appendix A.
@xref{Reference source code}.

@sp 1
@section What is coded

The LZMA stream includes literals, matches and repeated matches (matches
reusing a recently used distance). There are 7 different coding sequences:

@multitable @columnfractions .35 .14 .51
@headitem Bit sequence @tab Name @tab Description
@item 0 + byte @tab literal @tab literal byte
@item 1 + 0 + len + dis @tab match @tab distance-length pair
@item 1 + 1 + 0 + 0 @tab shortrep @tab 1 byte match at latest used
distance
@item 1 + 1 + 0 + 1 + len @tab rep0 @tab len bytes match at latest used
distance
@item 1 + 1 + 1 + 0 + len @tab rep1 @tab len bytes match at second
latest used distance
@item 1 + 1 + 1 + 1 + 0 + len @tab rep2 @tab len bytes match at third
latest used distance
@item 1 + 1 + 1 + 1 + 1 + len @tab rep3 @tab len bytes match at fourth
latest used distance
@end multitable

@sp 1
In the following tables, multibit sequences are coded in normal order,
from MSB to LSB, except where noted otherwise.

Lengths (the @samp{len} in the table above) are coded as follows:

@multitable @columnfractions .5 .5
@headitem Bit sequence @tab Description
@item 0 + 3 bits @tab lengths from 2 to 9
@item 1 + 0 + 3 bits @tab lengths from 10 to 17
@item 1 + 1 + 8 bits @tab lengths from 18 to 273
@end multitable

@sp 1
The coding of distances is a little more complicated, so I'll begin
explaining a simpler version of the encoding.

Imagine you need to code a number from 0 to @w{2^32 - 1}, and you want
to do it in a way that produces shorter codes for the smaller numbers.
You may first send the position of the most significant bit that is set
to 1, which you may find by making a bit scan from the left (from the
MSB). A position of 0 means that the number is 0 (no bit is set), 1
means the LSB is the first bit set (the number is 1), and 32 means the
MSB is set (i.e., the number is @w{>= 0x80000000}). Let's call this bit
position a "slot". Then, if slot is @w{> 1}, you send the remaining
@w{slot - 1} bits. Let's call these bits "direct_bits" because they are
coded directly by value instead of indirectly by position.

The inconvenient of this simple method is that it needs 6 bits to code
the slot, but it just uses 33 of the 64 possible values, wasting almost
half of the codes.

The intelligent trick of LZMA is that it encodes the position of the
most significant bit set, along with the value of the next bit, in the
same 6 bits that would take to encode the position alone. This seems to
need 66 slots (2 * position + next_bit), but for slots 0 and 1 there is
no next bit, so the number of needed slots is 64 (0 to 63).

The 6 bits representing this "slot number" are then context-coded. If
the distance is @w{>= 4}, the remaining bits are coded as follows.
@samp{direct_bits} is the amount of remaining bits (from 0 to 30) needed
to form a complete distance, and is calculated as @w{(slot >> 1) - 1}.
If a distance needs 6 or more direct_bits, the last 4 bits are coded
separately. The last piece (all the direct_bits for distances 4 to 127
or the last 4 bits for distances @w{>= 128}) is context-coded in reverse
order (from LSB to MSB). For distances @w{>= 128}, the
@w{@samp{direct_bits - 4}} part is coded with fixed 0.5 probability.

@multitable @columnfractions .5 .5
@headitem Bit sequence @tab Description
@item slot @tab distances from 0 to 3
@item slot + direct_bits @tab distances from 4 to 127
@item slot + (direct_bits - 4) + 4 bits @tab distances from 128 to
2^32 - 1
@end multitable

@sp 1
@section The coding contexts

These contexts (@samp{Bit_model} in the source), are integers or arrays
of integers representing the probability of the corresponding bit being 0.

The indices used in these arrays are:

@table @samp
@item state
A state machine (@samp{State} in the source) with 12 states (0 to 11),
coding the latest 2 to 4 types of sequences processed. The initial state
is 0.

@item pos_state
Value of the 2 least significant bits of the current position in the
decoded data.

@item literal_state
Value of the 3 most significant bits of the latest byte decoded.

@item len_state
Coded value of length @w{(length - 2)}, with a maximum of 3. The
resulting value is in the range 0 to 3.

@end table


In the following table, @samp{!literal} is any sequence except a literal
byte. @samp{rep} is any one of @samp{rep0}, @samp{rep1}, @samp{rep2} or
@samp{rep3}. The types of previous sequences corresponding to each state
are:

@multitable {State} {rep or (!literal, shortrep), literal, literal}
@headitem State @tab Types of previous sequences
@item  0 @tab literal, literal, literal
@item  1 @tab match, literal, literal
@item  2 @tab rep or (!literal, shortrep), literal, literal
@item  3 @tab literal, shortrep, literal, literal
@item  4 @tab match, literal
@item  5 @tab rep or (!literal, shortrep), literal
@item  6 @tab literal, shortrep, literal
@item  7 @tab literal, match
@item  8 @tab literal, rep
@item  9 @tab literal, shortrep
@item 10 @tab !literal, match
@item 11 @tab !literal, (rep or shortrep)
@end multitable

@sp 1
The contexts for decoding the type of coding sequence are:

@multitable @columnfractions .2 .4 .4
@headitem Name @tab Indices @tab Used when
@item bm_match @tab state, pos_state @tab sequence start
@item bm_rep @tab state @tab after sequence 1
@item bm_rep0 @tab state @tab after sequence 11
@item bm_rep1 @tab state @tab after sequence 111
@item bm_rep2 @tab state @tab after sequence 1111
@item bm_len @tab state, pos_state @tab after sequence 110
@end multitable

@sp 1
The contexts for decoding distances are:

@multitable @columnfractions .2 .4 .4
@headitem Name @tab Indices @tab Used when
@item bm_dis_slot @tab len_state, bit tree @tab distance start
@item bm_dis @tab reverse bit tree @tab after slots 4 to 13
@item bm_align @tab reverse bit tree @tab for distances >= 128, after
fixed probability bits
@end multitable

@sp 1
There are two separate sets of contexts for lengths (@samp{Len_model} in
the source). One for normal matches, the other for repeated matches. The
contexts in each Len_model are (see @samp{decode_len} in the source):

@multitable @columnfractions .2 .4 .4
@headitem Name @tab Indices @tab Used when
@item choice1 @tab none @tab length start
@item choice2 @tab none @tab after sequence 1
@item bm_low @tab pos_state, bit tree @tab after sequence 0
@item bm_mid @tab pos_state, bit tree @tab after sequence 10
@item bm_high @tab bit tree @tab after sequence 11
@end multitable

@sp 1
The context array @samp{bm_literal} is special. In principle it acts as
a normal bit tree context, the one selected by @samp{literal_state}. But
if the previous decoded byte was not a literal, two other bit tree
contexts are used depending on the value of each bit in
@samp{match_byte} (the byte at the latest used distance), until a bit is
decoded that is different from its corresponding bit in
@samp{match_byte}. After the first difference is found, the rest of the
byte is decoded using the normal bit tree context. (See
@samp{decode_matched} in the source).

@sp 1
@section The range decoder

The LZMA stream is consumed one byte at a time by the range decoder.
(See @samp{normalize} in the source). Every byte consumed produces a
variable number of decoded bits, depending on how well these bits agree
with their context. (See @samp{decode_bit} in the source).

The range decoder state consists of two unsigned 32-bit variables;
@code{range} (representing the most significant part of the range size
not yet decoded), and @code{code} (representing the current point within
@code{range}). @code{range} is initialized to @w{(2^32 - 1)}, and
@code{code} is initialized to 0.

The range encoder produces a first 0 byte that must be ignored by the
range decoder. This is done by shifting 5 bytes in the initialization of
@code{code} instead of 4. (See the @samp{Range_decoder} constructor in
the source).

@sp 1
@section Decoding the LZMA stream

After decoding the member header and obtaining the dictionary size, the
range decoder is initialized and then the LZMA decoder enters a loop
(See @samp{decode_member} in the source) where it invokes the range
decoder with the appropriate contexts to decode the different coding
sequences (matches, repeated matches, and literal bytes), until the "End
Of Stream" marker is decoded.


@node Trailing data
@chapter Extra data appended to the file
@cindex trailing data

Sometimes extra data are found appended to a lzip file after the last
member. Such trailing data may be:

@itemize @bullet
@item
Padding added to make the file size a multiple of some block size, for
example when writing to a tape. It is safe to append any amount of
padding zero bytes to a lzip file.

@item
Useful data added by the user; a cryptographically secure hash, a
description of file contents, etc. It is safe to append any amount of
text to a lzip file as long as none of the first four bytes of the text
match the corresponding byte in the string "LZIP", and the text does not
contain any zero bytes (null characters). Nonzero bytes and zero bytes
can't be safely mixed in trailing data.

@item
Garbage added by some not totally successful copy operation.

@item
Malicious data added to the file in order to make its total size and
hash value (for a chosen hash) coincide with those of another file.

@item
In rare cases, trailing data could be the corrupt header of another
member. In multimember or concatenated files the probability of
corruption happening in the magic bytes is 5 times smaller than the
probability of getting a false positive caused by the corruption of the
integrity information itself. Therefore it can be considered to be below
the noise level. Additionally, the test used by clzip to discriminate
trailing data from a corrupt header has a Hamming distance (HD) of 3,
and the 3 bit flips must happen in different magic bytes for the test to
fail. In any case, the option @samp{--trailing-error} guarantees that
any corrupt header will be detected.
@end itemize

Trailing data are in no way part of the lzip file format, but tools
reading lzip files are expected to behave as correctly and usefully as
possible in the presence of trailing data.

Trailing data can be safely ignored in most cases. In some cases, like
that of user-added data, they are expected to be ignored. In those cases
where a file containing trailing data must be rejected, the option
@samp{--trailing-error} can be used. @xref{--trailing-error}.


@node Examples
@chapter A small tutorial with examples
@cindex examples

WARNING! Even if clzip is bug-free, other causes may result in a corrupt
compressed file (bugs in the system libraries, memory errors, etc).
Therefore, if the data you are going to compress are important, give the
@samp{--keep} option to clzip and don't remove the original file until
you verify the compressed file with a command like
@w{@samp{clzip -cd file.lz | cmp file -}}.

@sp 1
@noindent
Example 1: Replace a regular file with its compressed version
@samp{file.lz} and show the compression ratio.

@example
clzip -v file
@end example

@sp 1
@noindent
Example 2: Like example 1 but the created @samp{file.lz} is multimember
with a member size of @w{1 MiB}. The compression ratio is not shown.

@example
clzip -b 1MiB file
@end example

@sp 1
@noindent
Example 3: Restore a regular file from its compressed version
@samp{file.lz}. If the operation is successful, @samp{file.lz} is
removed.

@example
clzip -d file.lz
@end example

@sp 1
@noindent
Example 4: Verify the integrity of the compressed file @samp{file.lz}
and show status.

@example
clzip -tv file.lz
@end example

@sp 1
@noindent
Example 5: Compress a whole device in /dev/sdc and send the output to
@samp{file.lz}.

@example
clzip -c /dev/sdc > file.lz
@end example

@sp 1
@anchor{concat-example}
@noindent
Example 6: The right way of concatenating the decompressed output of two
or more compressed files. @xref{Trailing data}.

@example
Don't do this
  cat file1.lz file2.lz file3.lz | clzip -d
Do this instead
  clzip -cd file1.lz file2.lz file3.lz
@end example

@sp 1
@noindent
Example 7: Decompress @samp{file.lz} partially until @w{10 KiB} of
decompressed data are produced.

@example
clzip -cd file.lz | dd bs=1024 count=10
@end example

@sp 1
@noindent
Example 8: Decompress @samp{file.lz} partially from decompressed byte
10000 to decompressed byte 15000 (5000 bytes are produced).

@example
clzip -cd file.lz | dd bs=1000 skip=10 count=5
@end example

@sp 1
@noindent
Example 9: Create a multivolume compressed tar archive with a volume
size of @w{1440 KiB}.

@example
tar -c some_directory | clzip -S 1440KiB -o volume_name
@end example

@sp 1
@noindent
Example 10: Extract a multivolume compressed tar archive.

@example
clzip -cd volume_name*.lz | tar -xf -
@end example

@sp 1
@noindent
Example 11: Create a multivolume compressed backup of a large database
file with a volume size of @w{650 MB}, where each volume is a
multimember file with a member size of @w{32 MiB}.

@example
clzip -b 32MiB -S 650MB big_db
@end example


@node Problems
@chapter Reporting bugs
@cindex bugs
@cindex getting help

There are probably bugs in clzip. There are certainly errors and
omissions in this manual. If you report them, they will get fixed. If
you don't, no one will ever know about them and they will remain unfixed
for all eternity, if not longer.

If you find a bug in clzip, please send electronic mail to
@email{lzip-bug@@nongnu.org}. Include the version number, which you can
find by running @w{@code{clzip --version}}.


@node Reference source code
@appendix Reference source code
@cindex reference source code

@verbatim
/*  Lzd - Educational decompressor for the lzip format
    Copyright (C) 2013-2018 Antonio Diaz Diaz.

    This program is free software. Redistribution and use in source and
    binary forms, with or without modification, are permitted provided
    that the following conditions are met:

    1. Redistributions of source code must retain the above copyright
    notice, this list of conditions and the following disclaimer.

    2. Redistributions in binary form must reproduce the above copyright
    notice, this list of conditions and the following disclaimer in the
    documentation and/or other materials provided with the distribution.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
*/
/*
    Exit status: 0 for a normal exit, 1 for environmental problems
    (file not found, invalid flags, I/O errors, etc), 2 to indicate a
    corrupt or invalid input file.
*/

#include <algorithm>
#include <cerrno>
#include <cstdio>
#include <cstdlib>
#include <cstring>
#include <stdint.h>
#include <unistd.h>
#if defined(__MSVCRT__) || defined(__OS2__) || defined(_MSC_VER)
#include <fcntl.h>
#include <io.h>
#endif


class State
  {
  int st;

public:
  enum { states = 12 };
  State() : st( 0 ) {}
  int operator()() const { return st; }
  bool is_char() const { return st < 7; }

  void set_char()
    {
    static const int next[states] = { 0, 0, 0, 0, 1, 2, 3, 4, 5, 6, 4, 5 };
    st = next[st];
    }
  void set_match()     { st = ( st < 7 ) ? 7 : 10; }
  void set_rep()       { st = ( st < 7 ) ? 8 : 11; }
  void set_short_rep() { st = ( st < 7 ) ? 9 : 11; }
  };


enum {
  min_dictionary_size = 1 << 12,
  max_dictionary_size = 1 << 29,
  literal_context_bits = 3,
  literal_pos_state_bits = 0,				// not used
  pos_state_bits = 2,
  pos_states = 1 << pos_state_bits,
  pos_state_mask = pos_states - 1,

  len_states = 4,
  dis_slot_bits = 6,
  start_dis_model = 4,
  end_dis_model = 14,
  modeled_distances = 1 << (end_dis_model / 2),		// 128
  dis_align_bits = 4,
  dis_align_size = 1 << dis_align_bits,

  len_low_bits = 3,
  len_mid_bits = 3,
  len_high_bits = 8,
  len_low_symbols = 1 << len_low_bits,
  len_mid_symbols = 1 << len_mid_bits,
  len_high_symbols = 1 << len_high_bits,
  max_len_symbols = len_low_symbols + len_mid_symbols + len_high_symbols,

  min_match_len = 2,					// must be 2

  bit_model_move_bits = 5,
  bit_model_total_bits = 11,
  bit_model_total = 1 << bit_model_total_bits };

struct Bit_model
  {
  int probability;
  Bit_model() : probability( bit_model_total / 2 ) {}
  };

struct Len_model
  {
  Bit_model choice1;
  Bit_model choice2;
  Bit_model bm_low[pos_states][len_low_symbols];
  Bit_model bm_mid[pos_states][len_mid_symbols];
  Bit_model bm_high[len_high_symbols];
  };


class CRC32
  {
  uint32_t data[256];		// Table of CRCs of all 8-bit messages.

public:
  CRC32()
    {
    for( unsigned n = 0; n < 256; ++n )
      {
      unsigned c = n;
      for( int k = 0; k < 8; ++k )
        { if( c & 1 ) c = 0xEDB88320U ^ ( c >> 1 ); else c >>= 1; }
      data[n] = c;
      }
    }

  void update_buf( uint32_t & crc, const uint8_t * const buffer,
                   const int size ) const
    {
    for( int i = 0; i < size; ++i )
      crc = data[(crc^buffer[i])&0xFF] ^ ( crc >> 8 );
    }
  };

const CRC32 crc32;


typedef uint8_t File_header[6];	// 0-3 magic, 4 version, 5 coded_dict_size

typedef uint8_t File_trailer[20];
			//  0-3  CRC32 of the uncompressed data
			//  4-11 size of the uncompressed data
			// 12-19 member size including header and trailer

class Range_decoder
  {
  uint32_t code;
  uint32_t range;

public:
  Range_decoder() : code( 0 ), range( 0xFFFFFFFFU )
    {
    for( int i = 0; i < 5; ++i ) code = (code << 8) | get_byte();
    }

  uint8_t get_byte() { return std::getc( stdin ); }

  unsigned decode( const int num_bits )
    {
    unsigned symbol = 0;
    for( int i = num_bits; i > 0; --i )
      {
      range >>= 1;
      symbol <<= 1;
      if( code >= range ) { code -= range; symbol |= 1; }
      if( range <= 0x00FFFFFFU )			// normalize
        { range <<= 8; code = (code << 8) | get_byte(); }
      }
    return symbol;
    }

  unsigned decode_bit( Bit_model & bm )
    {
    unsigned symbol;
    const uint32_t bound = ( range >> bit_model_total_bits ) * bm.probability;
    if( code < bound )
      {
      range = bound;
      bm.probability += (bit_model_total - bm.probability) >> bit_model_move_bits;
      symbol = 0;
      }
    else
      {
      range -= bound;
      code -= bound;
      bm.probability -= bm.probability >> bit_model_move_bits;
      symbol = 1;
      }
    if( range <= 0x00FFFFFFU )				// normalize
      { range <<= 8; code = (code << 8) | get_byte(); }
    return symbol;
    }

  unsigned decode_tree( Bit_model bm[], const int num_bits )
    {
    unsigned symbol = 1;
    for( int i = 0; i < num_bits; ++i )
      symbol = ( symbol << 1 ) | decode_bit( bm[symbol] );
    return symbol - (1 << num_bits);
    }

  unsigned decode_tree_reversed( Bit_model bm[], const int num_bits )
    {
    unsigned symbol = decode_tree( bm, num_bits );
    unsigned reversed_symbol = 0;
    for( int i = 0; i < num_bits; ++i )
      {
      reversed_symbol = ( reversed_symbol << 1 ) | ( symbol & 1 );
      symbol >>= 1;
      }
    return reversed_symbol;
    }

  unsigned decode_matched( Bit_model bm[], const unsigned match_byte )
    {
    unsigned symbol = 1;
    for( int i = 7; i >= 0; --i )
      {
      const unsigned match_bit = ( match_byte >> i ) & 1;
      const unsigned bit = decode_bit( bm[symbol+(match_bit<<8)+0x100] );
      symbol = ( symbol << 1 ) | bit;
      if( match_bit != bit )
        {
        while( symbol < 0x100 )
          symbol = ( symbol << 1 ) | decode_bit( bm[symbol] );
        break;
        }
      }
    return symbol & 0xFF;
    }

  unsigned decode_len( Len_model & lm, const int pos_state )
    {
    if( decode_bit( lm.choice1 ) == 0 )
      return decode_tree( lm.bm_low[pos_state], len_low_bits );
    if( decode_bit( lm.choice2 ) == 0 )
      return len_low_symbols +
             decode_tree( lm.bm_mid[pos_state], len_mid_bits );
    return len_low_symbols + len_mid_symbols +
           decode_tree( lm.bm_high, len_high_bits );
    }
  };


class LZ_decoder
  {
  unsigned long long partial_data_pos;
  Range_decoder rdec;
  const unsigned dictionary_size;
  uint8_t * const buffer;	// output buffer
  unsigned pos;			// current pos in buffer
  unsigned stream_pos;		// first byte not yet written to stdout
  uint32_t crc_;
  bool pos_wrapped;

  void flush_data();

  uint8_t peek( const unsigned distance ) const
    {
    if( pos > distance ) return buffer[pos - distance - 1];
    if( pos_wrapped ) return buffer[dictionary_size + pos - distance - 1];
    return 0;			// prev_byte of first byte
    }

  void put_byte( const uint8_t b )
    {
    buffer[pos] = b;
    if( ++pos >= dictionary_size ) flush_data();
    }

public:
  explicit LZ_decoder( const unsigned dict_size )
    :
    partial_data_pos( 0 ),
    dictionary_size( dict_size ),
    buffer( new uint8_t[dictionary_size] ),
    pos( 0 ),
    stream_pos( 0 ),
    crc_( 0xFFFFFFFFU ),
    pos_wrapped( false )
    {}

  ~LZ_decoder() { delete[] buffer; }

  unsigned crc() const { return crc_ ^ 0xFFFFFFFFU; }
  unsigned long long data_position() const { return partial_data_pos + pos; }

  bool decode_member();
  };


void LZ_decoder::flush_data()
  {
  if( pos > stream_pos )
    {
    const unsigned size = pos - stream_pos;
    crc32.update_buf( crc_, buffer + stream_pos, size );
    errno = 0;
    if( std::fwrite( buffer + stream_pos, 1, size, stdout ) != size )
      { std::fprintf( stderr, "Write error: %s\n", std::strerror( errno ) );
        std::exit( 1 ); }
    if( pos >= dictionary_size )
      { partial_data_pos += pos; pos = 0; pos_wrapped = true; }
    stream_pos = pos;
    }
  }


bool LZ_decoder::decode_member()		// Returns false if error
  {
  Bit_model bm_literal[1<<literal_context_bits][0x300];
  Bit_model bm_match[State::states][pos_states];
  Bit_model bm_rep[State::states];
  Bit_model bm_rep0[State::states];
  Bit_model bm_rep1[State::states];
  Bit_model bm_rep2[State::states];
  Bit_model bm_len[State::states][pos_states];
  Bit_model bm_dis_slot[len_states][1<<dis_slot_bits];
  Bit_model bm_dis[modeled_distances-end_dis_model+1];
  Bit_model bm_align[dis_align_size];
  Len_model match_len_model;
  Len_model rep_len_model;
  unsigned rep0 = 0;		// rep[0-3] latest four distances
  unsigned rep1 = 0;		// used for efficient coding of
  unsigned rep2 = 0;		// repeated distances
  unsigned rep3 = 0;
  State state;

  while( !std::feof( stdin ) && !std::ferror( stdin ) )
    {
    const int pos_state = data_position() & pos_state_mask;
    if( rdec.decode_bit( bm_match[state()][pos_state] ) == 0 )	// 1st bit
      {
      const uint8_t prev_byte = peek( 0 );
      const int literal_state = prev_byte >> ( 8 - literal_context_bits );
      Bit_model * const bm = bm_literal[literal_state];
      if( state.is_char() )
        put_byte( rdec.decode_tree( bm, 8 ) );
      else
        put_byte( rdec.decode_matched( bm, peek( rep0 ) ) );
      state.set_char();
      }
    else					// match or repeated match
      {
      int len;
      if( rdec.decode_bit( bm_rep[state()] ) != 0 )		// 2nd bit
        {
        if( rdec.decode_bit( bm_rep0[state()] ) == 0 )		// 3rd bit
          {
          if( rdec.decode_bit( bm_len[state()][pos_state] ) == 0 ) // 4th bit
            { state.set_short_rep(); put_byte( peek( rep0 ) ); continue; }
          }
        else
          {
          unsigned distance;
          if( rdec.decode_bit( bm_rep1[state()] ) == 0 )	// 4th bit
            distance = rep1;
          else
            {
            if( rdec.decode_bit( bm_rep2[state()] ) == 0 )	// 5th bit
              distance = rep2;
            else
              { distance = rep3; rep3 = rep2; }
            rep2 = rep1;
            }
          rep1 = rep0;
          rep0 = distance;
          }
        state.set_rep();
        len = min_match_len + rdec.decode_len( rep_len_model, pos_state );
        }
      else					// match
        {
        rep3 = rep2; rep2 = rep1; rep1 = rep0;
        len = min_match_len + rdec.decode_len( match_len_model, pos_state );
        const int len_state = std::min( len - min_match_len, len_states - 1 );
        rep0 = rdec.decode_tree( bm_dis_slot[len_state], dis_slot_bits );
        if( rep0 >= start_dis_model )
          {
          const unsigned dis_slot = rep0;
          const int direct_bits = ( dis_slot >> 1 ) - 1;
          rep0 = ( 2 | ( dis_slot & 1 ) ) << direct_bits;
          if( dis_slot < end_dis_model )
            rep0 += rdec.decode_tree_reversed( bm_dis + ( rep0 - dis_slot ),
                                               direct_bits );
          else
            {
            rep0 += rdec.decode( direct_bits - dis_align_bits ) << dis_align_bits;
            rep0 += rdec.decode_tree_reversed( bm_align, dis_align_bits );
            if( rep0 == 0xFFFFFFFFU )		// marker found
              {
              flush_data();
              return ( len == min_match_len );	// End Of Stream marker
              }
            }
          }
        state.set_match();
        if( rep0 >= dictionary_size || ( rep0 >= pos && !pos_wrapped ) )
          { flush_data(); return false; }
        }
      for( int i = 0; i < len; ++i ) put_byte( peek( rep0 ) );
      }
    }
  flush_data();
  return false;
  }


int main( const int argc, const char * const argv[] )
  {
  if( argc > 1 )
    {
    std::printf( "Lzd %s - Educational decompressor for the lzip format.\n",
                 PROGVERSION );
    std::printf( "Study the source to learn how a lzip decompressor works.\n"
                 "See the lzip manual for an explanation of the code.\n"
                 "It is not safe to use lzd for any real work.\n"
                 "\nUsage: %s < file.lz > file\n", argv[0] );
    std::printf( "Lzd decompresses from standard input to standard output.\n"
                 "\nCopyright (C) 2018 Antonio Diaz Diaz.\n"
                 "This is free software: you are free to change and redistribute it.\n"
                 "There is NO WARRANTY, to the extent permitted by law.\n"
                 "Report bugs to lzip-bug@nongnu.org\n"
                 "Lzd home page: http://www.nongnu.org/lzip/lzd.html\n" );
    return 0;
    }

#if defined(__MSVCRT__) || defined(__OS2__) || defined(_MSC_VER)
  setmode( fileno( stdin ), O_BINARY );
  setmode( fileno( stdout ), O_BINARY );
#endif

  for( bool first_member = true; ; first_member = false )
    {
    File_header header;				// verify header
    for( int i = 0; i < 6; ++i ) header[i] = std::getc( stdin );
    if( std::feof( stdin ) || std::memcmp( header, "LZIP\x01", 5 ) != 0 )
      {
      if( first_member )
        { std::fputs( "Bad magic number (file not in lzip format).\n", stderr );
          return 2; }
      break;
      }
    unsigned dict_size = 1 << ( header[5] & 0x1F );
    dict_size -= ( dict_size / 16 ) * ( ( header[5] >> 5 ) & 7 );
    if( dict_size < min_dictionary_size || dict_size > max_dictionary_size )
      { std::fputs( "Invalid dictionary size in member header.\n", stderr );
        return 2; }

    LZ_decoder decoder( dict_size );		// decode LZMA stream
    if( !decoder.decode_member() )
      { std::fputs( "Data error\n", stderr ); return 2; }

    File_trailer trailer;			// verify trailer
    for( int i = 0; i < 20; ++i ) trailer[i] = std::getc( stdin );
    unsigned crc = 0;
    for( int i = 3; i >= 0; --i ) { crc <<= 8; crc += trailer[i]; }
    unsigned long long data_size = 0;
    for( int i = 11; i >= 4; --i ) { data_size <<= 8; data_size += trailer[i]; }
    if( crc != decoder.crc() || data_size != decoder.data_position() )
      { std::fputs( "CRC error\n", stderr ); return 2; }
    }

  if( std::fclose( stdout ) != 0 )
    { std::fprintf( stderr, "Error closing stdout: %s\n", std::strerror( errno ) );
      return 1; }
  return 0;
  }
@end verbatim


@node Concept index
@unnumbered Concept index

@printindex cp

@bye