summaryrefslogtreecommitdiffstats
path: root/doc/lzip.texi
blob: 865a413cc95baa1d2f950379c1d82f023d6436d4 (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
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
\input texinfo @c -*-texinfo-*-
@c %**start of header
@setfilename lzip.info
@documentencoding ISO-8859-15
@settitle Lzip Manual
@finalout
@c %**end of header

@set UPDATED 1 March 2024
@set VERSION 1.24.1

@dircategory Compression
@direntry
* Lzip: (lzip).                 LZMA lossless data compressor
@end direntry


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

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

@contents
@end ifnothtml

@ifnottex
@node Top
@top

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

@menu
* Introduction::           Purpose and features of lzip
* Output::                 Meaning of lzip's output
* Invoking lzip::          Command-line interface
* Quality assurance::      Design, development, and testing of lzip
* Algorithm::              How lzip compresses the data
* File format::            Detailed format of the compressed file
* 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{} 2008-2024 Antonio Diaz Diaz.

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


@node Introduction
@chapter Introduction
@cindex introduction

@uref{http://www.nongnu.org/lzip/lzip.html,,Lzip}
is a lossless data compressor with a user interface similar to the one
of gzip or bzip2. Lzip uses a simplified form of the 'Lempel-Ziv-Markov
chain-Algorithm' (LZMA) stream format to maximize interoperability. The
maximum dictionary size is 512 MiB so that any lzip file can be decompressed
on 32-bit machines. Lzip provides accurate and robust 3-factor integrity
checking. 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. Lzip has been designed, written, and tested with great care to
replace gzip and bzip2 as the standard general-purpose compressed format for
Unix-like systems.

For compressing/decompressing large files on multiprocessor machines
@uref{http://www.nongnu.org/lzip/manual/plzip_manual.html,,plzip} can be
much faster than lzip at the cost of a slightly reduced compression ratio.
@ifnothtml
@xref{Top,plzip manual,,plzip}.
@end ifnothtml

For creation and manipulation of compressed tar archives
@uref{http://www.nongnu.org/lzip/manual/tarlz_manual.html,,tarlz} can be more
efficient than using tar and plzip because tarlz is able to keep the
alignment between tar members and lzip members.
@ifnothtml
@xref{Top,tarlz manual,,tarlz}.
@end ifnothtml

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 program
@uref{http://www.nongnu.org/lzip/manual/lziprecover_manual.html#Data-safety,,lziprecover}
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 lzip (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.

Lzip uses the same well-defined exit status values used by 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.

Lzip automatically uses for each file the largest dictionary size that does
not exceed neither the file size nor the limit given. 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
@option{-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, lzip replaces every file given in the command line
with a compressed version of itself, with the name "original_name.lz".
When decompressing, lzip 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 lzip
preserves the access and modification dates, permissions, and, if you have
appropriate privileges, ownership of the file just as @w{@samp{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).

Lzip is able to read from some types of non-regular files if either the
option @option{-c} or the option @option{-o} is specified.

Lzip refuses to read compressed data from a terminal or write compressed
data to a terminal, as this would be entirely incomprehensible and might
leave the terminal in an abnormal state.

Lzip correctly decompresses 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.

Lzip can produce multimember files, and lziprecover can safely recover the
undamaged members in case of file damage. Lzip 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.

Lzip 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 lzip's output
@cindex output

The output of lzip looks like this:

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

lzip -tvvv foo.lz
  foo.lz:  6.676:1, 14.98% ratio, 85.02% saved.  450560 out,  67493 in. 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
@w{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
Size of the input data. This is the uncompressed size when compressing, or
the compressed size when decompressing or testing. Note that lzip always
prints the uncompressed size before the compressed size when compressing,
decompressing, testing, or listing.

@item out
Size of the output data. This is the compressed size when compressing, or
the decompressed size when decompressing or testing.

@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 lzip
@chapter Invoking lzip
@cindex invoking
@cindex options
@cindex usage
@cindex version

The format for running lzip is:

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

@noindent
If no file names are specified, lzip compresses (or decompresses) from
standard input to standard output. A hyphen @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. Remember to
prepend @file{./} to any file name beginning with a hyphen, or use @samp{--}.

lzip supports the following
@uref{http://www.nongnu.org/arg-parser/manual/arg_parser_manual.html#Argument-syntax,,options}:
@ifnothtml
@xref{Argument syntax,,,arg_parser}.
@end ifnothtml

@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 lzip on the standard output and exit.
This version number should be included in all bug reports.

@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}. It is advisable
to keep members smaller than RAM size so that they can be repaired with
lziprecover in case of corruption. 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. (The
output consists of a sequence of independently compressed members). This
option (or @option{-o}) 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. @option{-c} overrides @option{-o}
and @option{-S}. @option{-c} has no effect when testing or listing.

@item -d
@itemx --decompress
Decompress the files specified. The integrity of the files specified is
checked. If a file does not exist, can't be opened, or the destination file
already exists and @option{--force} has not been specified, lzip continues
decompressing the rest of the files and exits with error status 1. If a file
fails to decompress, or is a terminal, lzip exits immediately with error
status 2 without decompressing the rest of the files. A terminal is
considered an uncompressed file, and therefore invalid.

@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
files specified. 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 @option{-v}, the dictionary
size, the number of members in the file, and the amount of trailing data (if
any) are also printed. With @option{-vv}, the positions and sizes of each
member in multimember files are also printed.

If any file is damaged, does not exist, can't be opened, or is not regular,
the final exit status is @w{> 0}. @option{-lq} can be used to check quickly
(without decompressing) the structural integrity of the files specified.
(Use @option{--test} to check the data integrity). @option{-alq}
additionally checks that none of the files specified 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}
If @option{-c} has not been also specified, write the (de)compressed output
to @var{file}, automatically creating any missing parent directories; keep
input files unchanged. If compressing several files, each file is compressed
independently. (The output consists of a sequence of independently
compressed members). This option (or @option{-c}) is needed when reading
from a named pipe (fifo) or from a device. @w{@option{-o -}} is equivalent
to @option{-c}. @option{-o} has no effect when testing or listing.

In order to keep backward compatibility with lzip versions prior to 1.22,
when compressing from standard input and no other file names are given, the
extension @samp{.lz} is appended to @var{file} unless it already ends in
@samp{.lz} or @samp{.tlz}. This feature will be removed in a future version
of lzip. Meanwhile, redirection may be used instead of @option{-o} to write
the compressed output to a file without the extension @samp{.lz} in its
name: @w{@samp{lzip < file > foo}}.

When compressing and splitting the output in volumes, @var{file} is used as
a prefix, and several files named @samp{@var{file}00001.lz},
@samp{@var{file}00002.lz}, etc, are created. In this case, only one input
file is allowed.

@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. Lzip uses for
each file the largest dictionary size that does not exceed neither the file
size nor 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. Dictionary sizes are quantized so that they can be coded in just one
byte (@pxref{coded-dict-size}). If the size specified does not match one of
the valid sizes, it is 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, and @option{-c} has not been also specified, 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 files specified, but don't decompress them. This
really performs a trial decompression and throws away the result. Use it
together with @option{-v} to see information about the files. If a file
fails the test, does not exist, can't be opened, or is a terminal, lzip
continues testing 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 @option{-v} options show the progress of (de)compression.

@item -0 .. -9
Compression level. Set the compression parameters (dictionary size and
match length limit) as shown in the table below. The default compression
level is @option{-6}, equivalent to @w{@option{-s8MiB -m36}}. Note that
@option{-9} can be much slower than @option{-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 options @option{--dictionary-size} and
@option{--match-length} directly to achieve optimal performance.

If several compression levels or @option{-s} or @option{-m} options are
given, the last setting is used. For example @w{@option{-9 -s64MiB}} is
equivalent to @w{@option{-s64MiB -m273}}

@multitable {Level} {Dictionary size (-s)} {Match length limit (-m)}
@item Level @tab Dictionary size (-s) @tab Match length limit (-m)
@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 --empty-error
Exit with error status 2 if any empty member is found in the input files.

@item --marking-error
Exit with error status 2 if the first LZMA byte is non-zero in any member of
the input files. This may be caused by data corruption or by deliberate
insertion of tracking information in the file. Use
@w{@samp{lziprecover --clear-marking}} to clear any such non-zero bytes.

@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 expressed in decimal,
hexadecimal, or octal (using the same syntax as integer constants in C++),
and 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)
@item R @tab ronnabyte  (10^27)       @tab | @tab Ri @tab robibyte  (2^90)
@item Q @tab quettabyte (10^30)       @tab | @tab Qi @tab quebibyte (2^100)
@end multitable

@sp 1
Exit status: 0 for a normal exit, 1 for environmental problems
(file not found, invalid command-line options, I/O errors, etc), 2 to
indicate a corrupt or invalid input file, 3 for an internal consistency
error (e.g., bug) which caused lzip 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 replace gzip
and bzip2 as the standard general-purpose compressed format for Unix-like
systems. This chapter describes the lessons learned from these previous
formats, and their application to the design of lzip. The lzip format
specification has been reviewed carefully and is believed to be free from
design errors.

@sp 1
@section Format design

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

Lziprecover is able to provide unique data recovery capabilities because the
lzip format is extraordinarily safe. The simple and safe design of the file
format complements the embedded error detection provided by the LZMA data
stream. 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
possible 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 optimal accuracy in the detection of
errors 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 several
million times more, resulting in a combined integrity checking optimally
accurate for any member size produced by lzip. Preliminary results suggest
that the lzip format is safe enough to be used in critical safety avionics
systems.

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 Hamming distance (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, then the header seems to be intact
(in spite of being corrupt) while the compressed blocks seem 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 or equal 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.
Additionally, 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

Our civilization depends critically on software; it had better be quality
software.@*
--- Bjarne Stroustrup

@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 method detects most false
positives for corruption. 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, data size, or
member size) in the member trailer.

@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 tested to check 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 adapts the dictionary size to the size of 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. @c key4_mask

@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 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". LZMA compression consists in describing the uncompressed data as a
succession of coding sequences from the set shown in Section @samp{What is
coded} (@pxref{what-is-coded}), and then encoding them using a range
encoder. For example, the option @option{-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.

Lzip currently implements two variants of the LZMA algorithm: fast
(used by option @option{-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) 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.

Lzip 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 bytes produced 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
During compression, lzip reads data in large blocks (one dictionary size at
a time). Therefore it may block for up to tens of seconds any process
feeding data to it through a pipe. This is normal. The blocking intervals
get longer with higher compression levels because dictionary size increases
(and compression speed decreases) with compression level.

@noindent
The ideas embodied in lzip are due to (at least) the following people:
Abraham Lempel and Jacob Ziv (for the LZ algorithm), Andrei 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 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 one or more independent "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 can
encode in compressed form up to @w{16 EiB - 1 byte} of uncompressed data.
The size of a multimember file is unlimited.

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.

@anchor{coded-dict-size}
@item DS (coded dictionary size, 1 byte)
The dictionary size is calculated by taking a power of 2 (the base size)
and subtracting 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 subtract
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)
Cyclic Redundancy Check (CRC) of the original uncompressed data.

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

@item Member size (8 bytes)
Total size of the member, including header and trailer. This field acts
as a distributed index, improves the checking of stream integrity, and
facilitates the safe recovery of undamaged members from multimember files.
Lzip limits the member size to @w{2 PiB} to prevent the data size field from
overflowing.

@end table


@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 finishes the LZMA stream with an "End Of Stream" (EOS) marker (the
distance-length pair @w{0xFFFFFFFFU, 2}), which in conjunction with the
@samp{member size} field in the member trailer allows the checking of stream
integrity. The EOS marker is the only LZMA marker allowed in lzip files. 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. This
simplified and marker-terminated form of the LZMA stream format has been
chosen to maximize interoperability and safety.

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 described individually.

It seems that the only way of describing the LZMA-302eos stream is to
describe 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 decompressor 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, included in appendix A. @xref{Reference source
code}. Lzd is written in C++11 and can be downloaded from the lzip download
directory.

@sp 1
@section What is coded

@anchor{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 most significant bit (MSB) to least significant bit (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 by
explaining a simpler version of the encoding.

Imagine you need to encode 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 encode 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}). Then, if the position is @w{>= 2}, you encode
the remaining @w{position - 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 encode the
position, 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 in what it calls a "slot"
the position of the most significant bit set, along with the value of the
next bit, using the same 6 bits that would take to encode the position
alone. This seems to need 66 slots (twice the number of positions), but for
positions 0 and 1 there is no next bit, so the number of slots needed 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 encoded as follows.
@samp{direct_bits} is the amount of remaining bits (from 1 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 encoded
separately. The last piece (all the direct_bits for distances 4 to 127
(slots 4 to 13), or the last 4 bits for distances @w{>= 128}
@w{(slot >= 14)}) is context-coded in reverse order (from LSB to MSB). For
distances @w{>= 128}, the @w{@samp{direct_bits - 4}} part is encoded 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 the current match length @w{(length - 2)}, with a maximum
of 3. The resulting value is in the range 0 to 3.

@end table


The types of previous sequences corresponding to each state are shown 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 last type in each line is the most recent.

@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 .35 .45
@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 .3 .5
@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:
@samp{range} (representing the most significant part of the range size
not yet decoded) and @samp{code} (representing the current point within
@samp{range}). @samp{range} is initialized to @w{2^32 - 1}, and
@samp{code} is initialized to 0.

The range encoder produces a first 0 byte that must be ignored by the
range decoder. (See the @samp{Range_decoder} constructor in the source).

@sp 1
@section Decoding and checking 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.

Once the "End Of Stream" marker has been decoded, the decompressor reads and
decodes the member trailer, and checks that the three integrity factors
stored there (CRC, data size, and member size) match those computed from the
data.


@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; an "End Of File" string (to check that the
file has not been truncated), 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 matches 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 lzip 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 @option{--trailing-error} guarantees that
any corrupt header is 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
@option{--trailing-error} can be used. @xref{--trailing-error}.


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

WARNING! Even if lzip 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
option @option{--keep} to lzip and don't remove the original file until you
check the compressed file with a command like
@w{@samp{lzip -cd file.lz | cmp file -}}. Most RAM errors happening during
compression can only be detected by comparing the compressed file with the
original because the corruption happens before lzip compresses the RAM
contents, resulting in a valid compressed file containing wrong data.

@sp 1
@noindent
Example 1: Extract all the files from archive @samp{foo.tar.lz}.

@example
  tar -xf foo.tar.lz
or
  lzip -cd foo.tar.lz | tar -xf -
@end example

@sp 1
@noindent
Example 2: Replace a regular file with its compressed version @samp{file.lz}
and show the compression ratio.

@example
lzip -v file
@end example

@sp 1
@noindent
Example 3: Like example 2 but the created @samp{file.lz} is multimember with
a member size of @w{1 MiB}. The compression ratio is not shown.

@example
lzip -b 1MiB file
@end example

@sp 1
@noindent
Example 4: Restore a regular file from its compressed version
@samp{file.lz}. If the operation is successful, @samp{file.lz} is removed.

@example
lzip -d file.lz
@end example

@sp 1
@noindent
Example 5: Check the integrity of the compressed file @samp{file.lz} and
show status.

@example
lzip -tv 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 | lzip -d -
Do this instead
  lzip -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
lzip -cd file.lz | dd bs=1024 count=10
@end example

@sp 1
@noindent
Example 8: Decompress @samp{file.lz} partially from decompressed byte at
offset 10000 to decompressed byte at offset 14999 (5000 bytes are produced).

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

@sp 1
@noindent
Example 9: Compress a whole device in /dev/sdc and send the output to
@samp{file.lz}.

@example
  lzip -c /dev/sdc > file.lz
or
  lzip /dev/sdc -o file.lz
@end example

@sp 1
@noindent
Example 10: Create a multivolume compressed tar archive with a volume size
of @w{1440 KiB}.

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

@sp 1
@noindent
Example 11: Extract a multivolume compressed tar archive.

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

@sp 1
@noindent
Example 12: 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
lzip -b 32MiB -S 650MB big_db
@end example


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

There are probably bugs in lzip. 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 lzip, please send electronic mail to
@email{lzip-bug@@nongnu.org}. Include the version number, which you can
find by running @w{@samp{lzip --version}}.


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

@verbatim
/* Lzd - Educational decompressor for the lzip format
   Copyright (C) 2013-2024 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 command-line options, 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 __DJGPP__
#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()
    {
    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;


enum { header_size = 6, trailer_size = 20 };
typedef uint8_t Lzip_header[header_size]; // 0-3 magic bytes
					  //   4 version
					  //   5 coded dictionary size
typedef uint8_t Lzip_trailer[trailer_size];
			//  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
  {
  unsigned long long member_pos;
  uint32_t code;
  uint32_t range;

public:
  Range_decoder()
    : member_pos( header_size ), code( 0 ), range( 0xFFFFFFFFU )
    {
    get_byte();			// discard first byte of the LZMA stream
    for( int i = 0; i < 4; ++i ) code = ( code << 8 ) | get_byte();
    }

  uint8_t get_byte() { ++member_pos; return std::getc( stdin ); }
  unsigned long long member_position() const { return member_pos; }

  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;
    }

  bool decode_bit( Bit_model & bm )
    {
    bool 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
      {
      code -= bound;
      range -= 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 bool match_bit = ( match_byte >> i ) & 1;
      const bool 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 min_match_len +
             decode_tree( lm.bm_low[pos_state], len_low_bits );
    if( decode_bit( lm.choice2 ) == 0 )
      return min_match_len + len_low_symbols +
             decode_tree( lm.bm_mid[pos_state], len_mid_bits );
    return min_match_len + 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; }
  uint8_t get_byte() { return rdec.get_byte(); }
  unsigned long long member_position() const
    { return rdec.member_position(); }

  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 );
    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()	// Return 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
      {
      // literal byte
      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();
      continue;
      }
    // 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 = rdec.decode_len( rep_len_model, pos_state );
      }
    else					// match
      {
      rep3 = rep2; rep2 = rep1; rep1 = rep0;
      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 > 2 || ( argc == 2 && std::strcmp( argv[1], "-d" ) != 0 ) )
    {
    std::printf(
      "Lzd %s - Educational decompressor for the lzip format.\n"
      "Study the source code to learn how a lzip decompressor works.\n"
      "See the lzip manual for an explanation of the code.\n"
      "\nUsage: %s [-d] < file.lz > file\n"
      "Lzd decompresses from standard input to standard output.\n"
      "\nCopyright (C) 2024 Antonio Diaz Diaz.\n"
      "License 2-clause BSD.\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",
      PROGVERSION, argv[0] );
    return 0;
    }

#if defined __MSVCRT__ || defined __OS2__ || defined __DJGPP__
  setmode( STDIN_FILENO, O_BINARY );
  setmode( STDOUT_FILENO, O_BINARY );
#endif

  for( bool first_member = true; ; first_member = false )
    {
    Lzip_header header;				// check header
    for( int i = 0; i < header_size; ++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;					// ignore trailing data
      }
    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; }

    Lzip_trailer trailer;			// check trailer
    for( int i = 0; i < trailer_size; ++i ) trailer[i] = decoder.get_byte();
    int retval = 0;
    unsigned crc = 0;
    for( int i = 3; i >= 0; --i ) crc = ( crc << 8 ) + trailer[i];
    if( crc != decoder.crc() )
      { std::fputs( "CRC mismatch\n", stderr ); retval = 2; }

    unsigned long long data_size = 0;
    for( int i = 11; i >= 4; --i )
      data_size = ( data_size << 8 ) + trailer[i];
    if( data_size != decoder.data_position() )
      { std::fputs( "Data size mismatch\n", stderr ); retval = 2; }

    unsigned long long member_size = 0;
    for( int i = 19; i >= 12; --i )
      member_size = ( member_size << 8 ) + trailer[i];
    if( member_size != decoder.member_position() )
      { std::fputs( "Member size mismatch\n", stderr ); retval = 2; }
    if( retval ) return retval;
    }

  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