/[pcre]/code/trunk/maint/README
ViewVC logotype

Diff of /code/trunk/maint/README

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

code/trunk/maintain/README revision 111 by ph10, Thu Mar 8 16:53:09 2007 UTC code/trunk/maint/README revision 129 by ph10, Tue Mar 20 16:33:54 2007 UTC
# Line 1  Line 1 
1  The files in the "maintain" directory of the PCRE source contain data, scripts,  MAINTENANCE README FOR PCRE
2  and programs that are used for the maintenance of PCRE, but do not form part of  ---------------------------
3  the PCRE distribution tarballs.  
4    The files in the "maint" directory of the PCRE source contain data, scripts,
5  Builducptable   A Perl script that creates the contents of the ucptable.h file  and programs that are used for the maintenance of PCRE, but which do not form
6                  from two Unicode data files, which themselves are downloaded  part of the PCRE distribution tarballs. This document describes these files and
7                  from the Unicode web site. Run this script in the "maintain"  also contains some notes for maintainers. Its contents are:
8                  directory.  
9      Files in the maint directory
10  Tech.Notes      Some notes about the internals of the PCRE code.    Updating to a new Unicode release
11      Preparing for a PCRE release
12  Unicode.tables  The files in this directory, Scripts.txt and UnicodeData.txt,    Making a PCRE release
13                  were downloaded from the Unicode web site. They contain    Long-term ideas (wish list)
14                  information about Unicode characters and scripts.  
15    
16  ucptest.c       A short C program for testing the Unicode property functions in  Files in the maint directory
17                  pcre_ucp_searchfuncs.c, mainly useful after rebuilding the  ----------------------------
18                  Unicode property table. Compile and run this in the "maintain"  
19                  directory.  Builducptable    A Perl script that creates the contents of the ucptable.h file
20                     from two Unicode data files, which themselves are downloaded
21  ucptestdata     A directory containing two files, testinput1 and testoutput1,                   from the Unicode web site. Run this script in the "maint"
22                  to use in conjunction with the ucptest program.                   directory.
23    
24  utf8.c          A short, freestanding C program for converting a Unicode code  ManyConfigTests  A shell script that runs "configure, make, test" a number of
25                  point into a sequence of bytes in the UTF-8 encoding, and vice                   times with different configuration settings.
26                  versa. If its argument is a hex number such as 0x1234, it  
27                  outputs a list of the equivalent UTF-8 bytes. If its argument  Unicode.tables   The files in this directory, Scripts.txt and UnicodeData.txt,
28                  is sequence of concatenated UTF-8 bytes (e.g. e188b4) it treats                   were downloaded from the Unicode web site. They contain
29                  them as a UTF-8 character and outputs the equivalent code point                   information about Unicode characters and scripts.
30                  in hex.  
31    ucptest.c        A short C program for testing the Unicode property functions
32                     in pcre_ucp_searchfuncs.c, mainly useful after rebuilding the
33                     Unicode property table. Compile and run this in the "maint"
34                     directory.
35    
36    ucptestdata      A directory containing two files, testinput1 and testoutput1,
37                     to use in conjunction with the ucptest program.
38    
39    utf8.c           A short, freestanding C program for converting a Unicode code
40                     point into a sequence of bytes in the UTF-8 encoding, and vice
41                     versa. If its argument is a hex number such as 0x1234, it
42                     outputs a list of the equivalent UTF-8 bytes. If its argument
43                     is sequence of concatenated UTF-8 bytes (e.g. e188b4) it
44                     treats them as a UTF-8 character and outputs the equivalent
45                     code point in hex.
46    
47    
48    Updating to a new Unicode release
49    ---------------------------------
50    
51  When there is a new release of Unicode, the files in Unicode.tables must be  When there is a new release of Unicode, the files in Unicode.tables must be
52  refreshed from the web site, and the Buildupctable script can then be run to  refreshed from the web site, and the Buildupctable script can then be run to
53  generate a new version of ucptable.h. The ucptest program can be used to check  generate a new version of ucptable.h. The ucptest program can be used to check
54  that the resulting table works properly, using the data files in ucptestdata to  that the resulting table works properly, using the data files in ucptestdata to
55  check a number of test characters.  check a number of test characters.
56    
57  ****  
58    Preparing for a PCRE release
59    ----------------------------
60    
61    This section contains a checklist of things that I consult before building a
62    distribution for a new release.
63    
64    . Ensure that the version number and version date are correct in configure.ac.
65    
66    . Run ./autogen.sh to ensure everything is up-to-date.
67    
68    . Compile and test with many different config options, and combinations of
69      options. The maint/ManyConfigTests script now encapsulates this testing.
70    
71    . Run perltest.pl on the test data for tests 1 and 4. The output should match
72      the PCRE test output, apart from the version identification at the top. The
73      other tests are not Perl-compatible (they use various special PCRE options).
74    
75    . Test on a number of different operating systems. In particular, at the moment
76      I can test on Solaris, using Sun's cc compiler (as a change from gcc). Adding
77      -xarch=v9 to the cc options does a 64-bit test, but it also needs -S 64 for
78      pcretest to increase the stack size for test 2. I also test on FreeBSD and
79      Linux (where I develop).
80    
81    . Test with valgrind by running "RunTest valgrind". There is also "RunGrepTest
82      valgrind", though that takes quite a long time.
83    
84    . It can also useful to test with Electric Fence, though the fact that it
85      grumbles for missing free() calls can be a nuisance. (A missing free() in
86      pcretest is hardly a big problem.) To build with EF, use:
87    
88        LIBS='/usr/lib/libefence.a -lpthread' with ./configure.
89    
90      Then all normal runs use it to check for buffer overflow. Also run everything
91      with:
92    
93        EF_PROTECT_BELOW=1 <whatever>
94    
95      because there have been problems with lookbehinds that looked too far.
96    
97    . Test with the emulated memmove() function by undefining HAVE_MEMMOVE and
98      HAVE_BCOPY in config.h.
99    
100    . Documentation: check AUTHORS, COPYING, ChangeLog (check date), INSTALL,
101      LICENCE, NEWS (check date), NON-UNIX-USE, and README. Many of these won't
102      need changing, but over the long term things do change.
103    
104    . Man pages: Check all man pages for \ not followed by e or f or " because
105      that indicates a markup error.
106    
107    
108    Making a PCRE release
109    ---------------------
110    
111    Run PrepareRelease and commit the files that it changes (by removing trailing
112    spaces). Then run "make dist" to create the tarballs and the zipball.
113    
114    Don't forget to update Freshmeat when the new release is out, and to tell
115    webmaster@pcre.org and the mailing list.
116    
117    
118    Future ideas (wish list)
119    ------------------------
120    
121    This section records a list of ideas so that they do not get forgotten. They
122    vary enormously in their usefulness and potential for implementation. Some are
123    very sensible; some are rather wacky. Some have been on this list for years;
124    others are relatively new.
125    
126    . Optimization
127    
128      There are always ideas for new optimizations so as to speed up pattern
129      matching. Most of them try to save work by recognizing a non-match without
130      having to scan all the possibilities. These are some that I've recorded:
131    
132      * /((A{0,5}){0,5}){0,5}(something complex)/ on a non-matching string is very
133        slow, though Perl is fast. Can we speed up somehow? Convert to {0,125}?
134        OTOH, this is pathological - the user could easily fix it.
135    
136      * Turn ={4} into ==== ? (for speed). I once did an experiment, and it seems
137        to have little effect, and maybe makes things worse.
138    
139      * "Ends with literal string" - note that a single character doesn't gain much
140        over the existing "required byte" (reqbyte) feature that just saves one
141        byte.
142    
143      * These probably need to go in study():
144    
145        o Remember an initial string rather than just 1 char?
146    
147        o A required byte from alternatives - not just the last char, but an
148          earlier one if common to all alternatives.
149    
150        o Minimum length of subject needed.
151    
152        o Friedl contains other ideas.
153    
154    . If Perl gets to a consistent state over the settings of capturing sub-
155      patterns inside repeats, see if we can match it. One example of the
156      difference is the matching of /(main(O)?)+/ against mainOmain, where PCRE
157      leaves $2 set. In Perl, it's unset. Changing this in PCRE will be very hard
158      because I think it needs much more state to be remembered.
159    
160    . Perl 6 will be a revolution. Is it a revolution too far for PCRE?
161    
162    . Unicode
163    
164      * Note that in Perl, \s matches \pZ and similarly for \d, \w and the POSIX
165        character classes. For the moment, I've chosen not to support this for
166        backward compatibility, for speed, and because it would be messy to
167        implement.
168    
169      * A different approach to Unicode might be to use a typedef to do everything
170        in unsigned shorts instead of unsigned chars. Actually, we'd have to have a
171        new typedef to distinguish data from bits of compiled pattern that are in
172        bytes, I think. There would need to be conversion functions in and out. I
173        don't think this is particularly trivial - and anyway, Unicode now has
174        characters that need more than 16 bits, so is this at all sensible?
175    
176      * There has been a request for direct support of 16-bit characters and
177        UTF-16. However, since Unicode is moving beyond purely 16-bit characters,
178        is this worth it at all? One possible way of handling 16-bit characters
179        would be to "load" them in the same way that UTF-8 characters are loaded.
180    
181    . Allow errorptr and erroroffset to be NULL. I don't like this idea.
182    
183    . Line endings:
184    
185      * Option to use NUL as a line terminator in subject strings. This could now
186        be done relatively easily since the extension to support LF, CR, and CRLF.
187        If this is done, a suitable option for pcregrep is also required.
188    
189    . Option to provide the pattern with a length instead of with a NUL terminator.
190      This probably affects quite a few places in the code.
191    
192    . Catch SIGSEGV for stack overflows?
193    
194    . "Cut" as described in Jeffrey Friedl's book, p364: \v and \V. The definitions
195      aren't yet clear enough for me. \v flushes saved states so that no
196      backtracking to anything earlier can happen; \V says "no more bumpalong", but
197      does it fail the current match? As described in the book, these aren't really
198      "cut" as in Prolog, are they? NOTE: (a) PCRE once had "cut", but it was
199      removed when atomic groups were introduced. (b) Perl 5.10 has some (*PRUNE)
200      features -- see below.
201    
202    . A feature to suspend a match via a callout was once requested.
203    
204    . Option to convert results into character offsets and character lengths.
205    
206    . Option for pcregrep to scan only the start of a file. I am not keen - this is
207      the job of "head".
208    
209    . A (non-Unix) user wanted pcregrep options to (a) list a file name just once,
210      preceded by a blank line, instead of adding it to every matched line, and (b)
211      support --outputfile=name.
212    
213    . Consider making UTF-8 and UCP the default for PCRE n.0 for some n > 7.
214    
215    . Add a user pointer to pcre_malloc/free functions -- some option would be
216      needed to retain backward compatibility.
217    
218    . Define a union for the results from pcre_fullinfo().
219    
220    . Provide a "random access to the subject" facility so that the way in which it
221      is stored is independent of PCRE. For efficiency, it probably isn't possible
222      to switch this dynamically. It would have to be specified when PCRE was
223      compiled. PCRE would then call a function every time it wanted a character.
224    
225    . There are new (*PRUNE) facilities in Perl 5.10, some of which it might be
226      relatively easy to implement.
227    
228    . Also in Perl 5.10 are relative subroutine references (?&-1) and (?&+1) which
229      I didn't know about when I added some 5.10 features for PCRE 7.0. What about
230      (?(-1)... as a condition? That's an obvious extension, even if Perl 5.10
231      doesn't have it.
232    
233    . Wild thought: the ability to compile from PCRE's internal byte code to a real
234      FSM and a very fast (third) matcher to process the result. There would be
235      even more restrictions than for pcre_dfa_exec(), however. This is not easy.
236    
237    . Should pcretest have some private locale data, to avoid relying on the
238      available locales for the test data, since different OS have different ideas?
239      This won't be as thorough a test, but perhaps that doesn't really matter.
240    
241    . pcregrep: add -rs for a sorted recurse? Having to store file names and sort
242      them will of course slow it down.
243    
244    . Re-arrange test 2: take out the link-size dependent stuff for a separate test
245      that is run only when the link size *is* 2; leave in some non-numbered
246      debugging tests using the new /Z feature.
247    
248    . Stan Switzer's goto replacement for longjmp, which is apparently very slow on
249      OS-X. This is used when stack recursion is disabled. It would be worth doing
250      some timing tests on other OS.
251    
252    . Someone suggested --disable-callout to save code space when callouts are
253      never wanted. This seems rather marginal.
254    
255    Philip Hazel
256    Email local part: ph10
257    Email domain: cam.ac.uk
258    Last updated: 20 March 2007

Legend:
Removed from v.111  
changed lines
  Added in v.129

  ViewVC Help
Powered by ViewVC 1.1.5