2000-09-26 Bruno Haible <bruno@linuix.math.u-bordeaux.fr> * gperf-2.7.2 released. * doc/gperf.texi: Add a second bug report address <gperf-bugs@lists.sourceforge.net>. * README: Updated. 2000-08-28 Bruno Haible <bruno@linuix.math.u-bordeaux.fr> * lib/getopt.h (struct option): Use "const" also when compiling in C++ mode. Avoids warnings from Sun CC and HP-UX aCC. * doc/Makefile.in (docdir): Change from $(datadir)/doc/@PACKAGE@ to $(prefix)/doc/@PACKAGE@, following the newest GNU standards. 2000-08-20 Bruno Haible <bruno@linuix.math.u-bordeaux.fr> * src/version.cc: Bump version number to 2.7.2. * doc/gperf.texi: Likewise. * doc/texinfo.tex: Update to post-texinfo-4.0 version. @code in the title page now chooses a larger font. The overall layout of the text is denser. * AUTHORS: New file. * tests/Makefile.in (all): Add check-lang-utf8 and check-lang-ucs2. (check-lang-utf8, check-lang-ucs2): New targets. (clean): Remove lu8out and lu2out. * tests/lang-utf8.gperf, tests/lang-utf8.exp: New files. * tests/lang-ucs2.gperf, tests/test2.c, tests/lang-ucs2.in, tests/lang-ucs2.exp: New files. Allow the use of embedded NULs in keys. * lib/hash.h (hashpjw): Add a length argument. * lib/hash.cc (hashpjw): Likewise. Don't stop when encountering a NUL character. * src/hash-table.h (Hash_Table constructor): Add ignore_len argument. (Hash_Table::ignore_length): New field. (Hash_Table::insert): Renamed from Hash_Table::operator(). Remove ignore_length argument. * src/hash-table.cc (NIL): Remove macro. (Hash_Table constructor): Add ignore_len argument. Use it to initialize ignore_length. (Hash_Table destructor): Specify explicit length of char_set and key. (Hash_Table::insert): Renamed from Hash_Table::operator(). Remove ignore_length argument. Pass explicit length to hashpjw. Compare char_set using memcmp, not strcmp. * src/list-node.h (List_Node): Rename field length to key_length. New field char_set_length. (List_Node constructor): Accept key and rest, not the entire line. * src/list-node.cc (List_Node constructor): Accept key and rest, not the entire line. Don't NUL terminate key and char_set. Specify explicit length of key. Initialize char_set_length field. * src/key-list.cc: Include <ctype.h>. (parse_line): New function. (Key_List::read_keys): Call parse_line instead of new List_Node. Pass option[NOLENGTH] to Hash_Table constructor, not Hash_Table::insert. Specify explicit length of key and char_set. (Key_List::get_occurrence): Use explicit length of char_set. (Key_List::set_determined): Likewise. (Key_List::already_determined): Likewise. (output_string): Add length argument. Output unprintable characters using octal escape sequence. (output_keyword_entry): Use explicit length of key. (Key_List::output_lookup_array): Specify explicit length of key. (output_switch_case): Likewise. (Key_List::dump): Likewise. * src/gen-perf.h (Gen_Perf::compute_disjoint_union): Add two length arguments. * src/gen-perf.cc (Gen_Perf::compute_disjoint_union): Likewise. Don't stop when encountering NUL characters. Don't NUL terminate the result. (Gen_Perf::hash): Use explicit length of char_set. (Gen_Perf::change): Specify explicit length of key. * doc/gperf.texi: Document it. * doc/help2man: New file, help2man version 1.022. * Makefile.devel (all): Add doc/gperf.1. (doc/gperf.1): New target. * doc/gperf.1: Automatically generated. * mkinstalldirs: New file, from automake-1.4, grep-2.4.1, guile-1.4, libtool-1.3.3, make-3.79.1, tar-1.13. * src/Makefile.in (MKINSTALLDIRS): New variable. (install, installdirs): Use it instead of mkdir. * doc/Makefile.in (MKINSTALLDIRS): New variable. (install, installdirs): Use it instead of mkdir. * INSTALL: Update. 2000-08-19 Bruno Haible <bruno@linuix.math.u-bordeaux.fr> * src/key-list.cc (Output_Compare_Memcmp): New class. (Key_List::output_lookup_function): When option -l is given, use memcmp instead of strcmp or strncmp. * doc/gperf.texi: The bug report address is <bug-gnu-utils@gnu.org>. The download address is ftp.gnu.org. Remove mention of -a and -g options (now nops). Explain effect of -c option. * doc/configure.in (PACKAGE): New variable. * doc/Makefile.in (datadir, docdir): New variables. (dvidir, htmldir): Change values. (install, installdirs): Update. * src/configure.in: Rename cache variable gp_cxx_throw_decl to gp_cv_cxx_throw_decl. * src/key-list.cc (Key_List::output_hash_function): When outputting __inline, take advantage of C++ compilers which have inline. * src/key-list.cc (Output_Compare_Strncmp::output_comparison): After the call to strncmp, verify that expr2 is not longer than `len'. Reported by Carlo Wood <carlo@runaway.xs4all.nl>. * src/key-list.cc (Key_List::output_lookup_function_body): Avoid emitting the loop for dealing with duplicates if total_duplicates == 0. * src/key-list.cc (Key_List::read_keys): Don't accept an empty key. * src/Makefile.in (install, installdirs, uninstall): Respect $(DESTDIR). * doc/Makefile.in (install, installdirs, uninstall): Likewise. * src/options.cc (Options::print_options): Escape the arguments which contain special characters. * tests/c-parse.gperf: Updated from gcc-2.95.2/gcc/c-parse.gperf. * tests/objc.gperf: New file, from gcc-2.95.2/gcc/objc/objc.gperf. * tests/chill.gperf: New file, from gcc-2.95.2/gcc/ch/gperf. * tests/cplusplus.gperf: New file, from gcc-2.95.2/gcc/cp/gxx.gperf. * tests/gplus.gperf: Remove file. * tests/java.gperf: New file, from gcc-2.95.2/gcc/java/keyword.gperf. * tests/Makefile: Check them all. * tests/c-parse.exp: Renamed from tests/test-1.exp. * tests/modula2.exp: Renamed from tests/test-2.exp. * tests/cplusplus.exp: Renamed from tests/test-3.exp. * tests/gpc.exp: Renamed from tests/test-5.exp. * src/key-list.cc (output_switch_case): Add trailing semicolon to lengthptr assignment line. Fixes bug with -D and -S. From Reini Urban <rurban@sbox.tu-graz.ac.at>. Also reported by David Hunter. * tests/Makefile.in (check-lang-syntax): Perform each test with -D once without and once with duplicates. * src/key-list.cc (output_keyword_entry): Avoid outputting a struct initializer of the form {"key",}. * src/iterator.cc: Don't include <stream.h>. From Michael Deutschmann <ldeutsch@mail.netshop.net>. * tests/Makefile.in (VALIDATE, check-lang-syntax): Use $(srcdir) where appropriate. Reported by Andreas Schwab <schwab@issan.informatik.uni-dortmund.de>. * tests/validate: Don't run -traditional tests by default. * src/main.cc (main): Check for write error on stdout before returning. * src/Makefile.in (LDFLAGS): New definition, to catch the value given at configure time. Make the structure initializers customizable. Based on a patch by Kaveh R. Ghazi <ghazi@caip.rutgers.edu>. * src/options.h (Options::get_initializer_suffix, Options::initializer_suffix): New declarations. * src/options.icc (Options::get_initializer_suffix): New function. * src/options.cc (DEFAULT_INITIALIZER_SUFFIX): New constant. (Options::initializer_suffix): New variable. (Options::short_usage): Document option "-F". (Options::long_usage): Document option "-F". (Options constructor): Initialize initializer_suffix. (Options destructor): Dump initializer_suffix. (long_options): Add option "-F". (Options::operator()): Accept option "-F". Sets initializer_suffix. * src/key-list.cc (output_keyword_blank_entries): Output initializer_suffix. * doc/gperf.texi: Document option "-F". * COPYING: Replace with GPL version 2 (with new FSF address and Y2K safe year format). * doc/gpl.texinfo: New file. * doc/gperf.texi: Document it. * doc/Makefile.in (gperf.info, gperf.dvi, gperf.html, gperf_toc.html): Update dependencies. * doc/Makefile.in (MAKEINFO): Unset LANG while running makeinfo. 1998-05-20 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * doc/Makefile.in (gperf.dvi, clean): Remove gperf.cps. (install, installdirs, uninstall): Don't install gperf.dvi. The info and HTML documentations are sufficient for on-line use, and users who wish to print the documentation (in PS or DVI format) can do this directly off the source distribution. (DVIPS): Use "-D600" instead of "-Pljfour", for portability. 1998-05-20 Akim Demaille <demaille@inf.enst.fr> * doc/gperf.texi: Many modifications: (Output Format): Declare `hash' and `in_word_set' as functions. (Concept Index): New section. (Title page): Use standard presentation. (Top): Use @top instead of @unnumbered so that automatic master update works. (Motivation): Avoid spaces in @var. (Options): Use the standard name ``Invoking ...''. (Options): Declare also the long form of the options. (Options): Remove redundant @itemize when @table is used. 1998-05-08 Andreas Schwab <schwab@issan.informatik.uni-dortmund.de> * aclocal.m4 (CL_PROG_INSTALL): Set cl_cv_path_install, not ac_cv_path_install. Sat May 2 13:20:54 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * gperf-2.7 released. Sat May 2 12:31:51 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * src/version.cc (version_string): Remove the "(C++ version)" suffix. It's redundant: the early C versions of gperf are called cperf. Reported by Karl Berry. * src/option.cc (Options::operator()): Trim the output of "gperf -v". Thu Apr 16 13:22:16 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * lib/Makefile.in, src/Makefile.in: Don't use $(TARGET_ARCH). Solaris "make" sets it to a value not understood by "cc". Wed Apr 15 23:52:14 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * lib/Makefile.in, src/Makefile.in: Don't use implicit rules. Don't use $<. AIX "make" and OSF/1 "make" have problems with both. * src/gen-perf.cc, src/key-list.cc: Cast free() argument to char*, otherwise it doesn't compile on SunOS 4. * src/key-list.h: Declare structs outside of other declarations, needed for OSF/1 cxx 5.5. * lib/getopt.h: Use prototypes if __STDC__ || __cplusplus. Don't give a prototype for getopt(), to avoid error on SunOS 4. * lib/getopt.c: Declare strncmp, to avoid warnings. Tue Apr 14 23:24:07 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * lib/GetOpt.{h,cc}: Remove files. * lib/getopt.{h,c}, lib/getopt1.c: New files, from GNU libc. * lib/configure.in (AC_INIT): Search for hash.cc, not GetOpt.cc. * lib/Makefile.in (OBJECTS): Remove GetOpt.o, add getopt.o, getopt1.o. (getopt.o, getopt1.o, hash.o): Use explicit building rules. Some "make"s don't support to have both implicit rules for "%.o : %.c" and "%.o : %.cc" in the same Makefile. * lib/hash.{h,cc}: Remove #pragma; there are no templates here. * src/option.h (Options::usage): Remove. (Options::short_usage, Options::long_usage): Declare. * src/option.cc (Options::usage): Remove. (Options::short_usage, Options::long_usage): New functions. (long_options): New array. (Options::operator()): Use getopt_long instead of GetOpt::operator(), change all references to GetOpt members. * src/std-err.{h,cc}: Remove files. * src/gen-perf.cc, src/key-list.cc, list-node.cc, new.cc, options.cc: Call fprintf(stderr) instead of Std_Err::report_error(). * src/key-list.h, src/list-node.h, src/options.h: Don't use class Std_Err any more. * src/option.cc (program_name): New variable. * src/Makefile.in: Remove STD_ERR_H. (OBJECTS): Remove std-err.o. Mon Mar 23 01:03:35 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * aclocal.m4, {lib,src,tests}/configure.in: Remove CL_CC_WORKS and CL_CXX_WORKS, already contained in autoconf 2.12. * src/gen-perf.cc, src/key-list.cc: Move some code from Gen_Perf::Gen_Perf() to Key_List::output(). * src/Makefile.in: Update dependencies. * src/options.{h,cc}: Remove option "-p". * src/key-list.cc (Key_List::set_output_types): Rewrite. (default_array_type, default_return_type): Remove. * src/key-list.cc: Adjust "const" handling. + With option "-t" [TYPE], don't emit wrong code if there is no space before the struct's opening brace. Sun Mar 22 16:59:15 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * src/key-list.{h,cc}: Completely reorganized the output routines. Rewrote from scratch the output_switch() function. Use classes (Output_Constants, Output_Expr, Output_Compare) for abstraction. In particular: + Don't emit trailing whitespace and spurious blank lines. + Adjust indentation of the arrays. + Don't emit commas at the end of array initializers and struct initializers. + With option "-l" [LENTABLE], compare the length before fetching the word from memory. + With option "-S" [SWITCH], emit the comparison code just once, not once in every switch statement. + With option "-S" [SWITCH], choose the right switch statement through a binary search, not a linear search. + With option "-S" [SWITCH], emit straightforward comparisons instead of switch statements with just one "case" label. + With options "-S -p -t" [SWITCH, POINTER, TYPE], don't emit spurious empty elements at the beginning of the wordlist array. + With option "-D" [DUP] and not option "-S" [SWITCH], if there is no more room for duplicate entries in the lookup array, don't call `assert (i != 0)'. Instead, make the array larger :-) + With option "-D" [DUP], if there are no duplicates, don't automatically fall back to the non-"-D" algorithm. If the user wants the non-"-D" algorithm, he can just not specify "-D". + With option "-D" [DUP] and either options "-p -t" [POINTER, TYPE] or not option "-S" [SWITCH], don't emit spurious empty elements at the beginning of the wordlist array. + With option "-D" [DUP], simplify the detection and processing of duplicate entries in the lookup array. + With options "-D -l" [DUP, LENTABLE] and not option "-S" [SWITCH], don't forget to emit the lengthtable array. + With options "-D -l -S" [DUP, LENTABLE, SWITCH], don't forget to compare the lengths before comparing the strings. * src/gen-perf.cc: No need to include <assert.h>. * src/options.cc: Likewise. * src/options.cc: Don't use `errno' after freopen failed. * src/std-err.cc: `report_error' doesn't call strerror(errno) any more. No need to include <string.h> and <errno.h>. * tests/Makefile.in (check-*): Any difference between .exp and .out is a failure. Don't ignore whitespace differences. * tests/Makefile.in (check-lang-syntax): Add some more checks. Fri Mar 20 00:54:54 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * tests/jscript.gperf: Renamed from tests/javascript.gperf, because of Minix and SVR2 14-character filename limit. * src/key-list.cc (output_string): New function. (Key_List::output_switch, Key_List::output_keyword_table): Call it. * src/options.{h,icc,cc} (get_wordlist_name): New function. Add option -W. * src/key-list.cc (Key_List::output_switch, Key_List::output_keyword_table, Key_List::output_lookup_function): Use it. Patch from William Bader <wbader@CSEE.Lehigh.Edu>. * src/version.cc: Bump version number directly from 2.5 to 2.7, because Schmidt's last release from 1991 carries version number 2.6. Tue Jul 30 00:02:39 1991 Douglas C. Schmidt (schmidt at net4.ics.uci.edu) * Fixed a small bug in the Key_List::output_keyword_table routine that caused an extra newline to be printed if there where no leading blank entries... (who cares, right?!) Mon Jul 29 22:05:40 1991 Douglas C. Schmidt (schmidt at net4.ics.uci.edu) * Modified the handling of the -E (emit enums rather than #defines) option in conjunction with the -G option. Now, if -G and -E are given the enums are generated outside the lookup function, rather than within it! Mon Apr 8 18:17:04 1991 Doug Schmidt (schmidt at net4.ics.uci.edu) * Yucko, there was a bug in the handling of -c (and of course the new -I command in key-list.cc). Apparently when I added the super-duper hack that provided support for duplicate keys I forgot to update the strcmp output... Mon Mar 9 02:19:04 1998 Bruno Haible <bruno@linuix.mathematik.uni-karlsruhe.de> * Moved the documentation to doc/, put the stuff borrowed from libg++ into lib/. * Rewrote all Makefile.in's for better compliance with GNU standards. * Autoconf based configuration. Rewrote all configure.in's. Added aclocal.m4, with macros from CLISP and CLN. Added Makefile.devel. * src/depend: Removed. Dependencies are now in src/Makefile.in. * src/bool-array.icc: New file, contains inline functions, from both src/bool-array.h and src/bool-array.cc. * src/options.icc: New file, contains inline functions, from both src/options.h and src/options.cc. * src/read-line.icc: New file, contains inline functions, from both src/read-line.h and src/read-line.cc. * src/bool-array.h: Don't include <std.h>. * src/bool-array.cc: Include <string.h>. * src/gen-perf.cc: No need to include <ctype.h>. Don't include <_G_config.h>. * src/hash-table.cc: Don't include <std.h> and <builtin.h>. Include <string.h> and lib/hash.h instead. * src/iterator.cc: Don't include <std.h>. * src/key-list.cc: Don't include <builtin.h>. Include <string.h> and <stdlib.h> instead. * src/list-node.cc: Don't include <std.h>. Include <stdlib.h> instead. Remove `index' hack. * src/main.cc: Don't include <_G_config.h>. * src/new.cc: Don't include <std.h>. Include <stdlib.h> instead. * src/options.cc: Don't include <builtin.h>. Include <string.h> and <stdlib.h> instead. * src/read-line.cc: Don't include <builtin.h>. Include <stdlib.h> instead. * src/std-err.cc: Don't include <std.h>. Include <string.h> (for Irix). * src/vectors.h: No need to include <stdio.h>. * src/version.cc: No need to include <stdio.h>. * src/bool-array.h: Change `STORAGE_TYPE' from int to unsigned int. * src/bool-array.{h,cc}: Change type of `Bool_Array::size' from int to unsigned int. * src/bool-array.{h,cc}: Change type of `Bool_Array::init' argument from STORAGE_TYPE to unsigned int. * src/gen-perf.{h,cc}: Change two `Gen_Perf::compute_disjoint_union' argument types from `char *' to `const char *'. * src/iterator.h: Change type of `Iterator::str' and argument of `Iterator::Iterator' from `char *' to `const char *'. * src/iterator.cc: Cast to `unsigned char' before calling `isdigit'. * src/key-list.{h,cc}: Change type of `Key_List::array_type', `Key_List::return_type', `Key_List::struct_tag', `Key_List::include_src', `default_array_type', `default_return_type' and return type of `Key_List::get_array_type', `Key_List::get_special_input', `Key_List::save_include_src' from `char *' to `const char *'. * src/key-list.cc: Change "pretty gross" assignment. * src/key-list.cc: Don't use `alloca', HP-UX CC lacks it. * lib/GetOpt.cc: Likewise. * src/key-list.cc (merge): Use iteration instead of recursion. * src/list-node.{h,cc}: Change type of `List_Node::key', `List_Node::rest', `List_Node::char_set' from `char *' to `const char *'. * src/new.cc: Don't use BUFSIZ. Conditionalize the throw() declaration. * src/read-line.h: Don't use BUFSIZ. * src/read-line.cc: Make CHUNK_SIZE a constant, and use allocate the buffers on the stack by default. Use memcpy for copying buffers. Include <string.h>. * src/read-line.icc (get_line): Use iteration instead of tail recursion. Don't call ungetc(EOF,stdin). * src/std-err.{h,cc}: Change type of `Std_Err::program_name' and of argument `Std_Err::report_error' from `char *' to `const char *'. * src/std-err.cc: `report_error' doesn't call `exit' any more. All callers changed to do that themselves. * src/trace.h: Make constructor/destructor calls non-inline. * src/key-list.cc (output_hash_function): If option[CPLUSPLUS], always make the hash function inline. (output): Declare the hash function inline, with the right name. * src/options.{h,cc}, src/gen-perf.cc, src/key-list.cc: Remove options -g, making it on by default. Remove option -a. Instead, introduce "-L KR-C", "-L C", "-L ANSI-C", "-L C++". * src/options.{h,cc}, src/key-list.cc: Add option -I. * src/key-list.cc: Don't emit "const" if compiling in mode "-L KR-C". * src/key-list.cc: Don't emit a comma at the end of an enum list. * src/main.cc: Remove COUNT_TIME code. * src/vectors.h, src/key-list.cc, src/options.cc, src/list-node.cc: ALPHA_SIZE defaults to 256 now. Add option -7. * tests/javascript.gperf: New file. * tests/jstest*.gperf, tests/validate: New tests. Sat Jan 31 01:38:11 1998 Alexandre Oliva <oliva@dcc.unicamp.br> * src/Makefile.in ($(TARGETPROG)): Add $(CFLAGS). Wed Jan 28 01:56:00 1998 Manfred Hollstein <manfred@s-direktnet.de> * configure.in (package_makefile_rules_frag): New and redirect stderr from ${srcdir}/config.shared to ${package_makefile_rules_frag}. * src/configure.in: Ditto. * tests/configure.in: Ditto. Fri Jan 23 08:00:41 1998 H.J. Lu (hjl@gnu.org) * gperf.texi (@ichapter): Changed to @chapter. Wed Jan 14 09:16:48 1998 H.J. Lu (hjl@gnu.org) * src/key-list.cc, src/key-list.h (Key_List::strcspn): Don't define if strcspn is defined. Fri Jan 24 13:23:47 1997 Mike Stump <mrs@cygnus.com> * src/new.cc (operator delete): Add the exception specification. Mon Feb 5 19:29:16 1996 Per Bothner <bothner@kalessin.cygnus.com> * src/read-line.cc (Read_Line::readln_aux): Handle EOF if last line has a length which is an exact multiple of CHUNK_SIZE. (Used to throw away the line's contents.) From Bruno Haible <haible@ilog.ilog.fr>. * src/Makefile.in ($(TARGETPROG)): Add -lm to link line. Tue Jun 11 13:43:50 1996 Brendan Kehoe <brendan@lisa.cygnus.com> * src/list-node.cc (List_Node): Reorder init of nodes to match declaration order. * src/hash-table.cc (Hash_Table): Likewise. Tue Oct 10 16:37:28 1995 Mike Stump <mrs@cygnus.com> * src/new.cc: Since malloc/delete are not paired, we cannot call free. Wed Jan 4 12:40:14 1995 Per Bothner <bothner@kalessin.cygnus.com> * src/Makefile.in ($(TARGETPROG)): Link with $(LDFLAGS). Patch from John Interrante <interran@uluru.stanford.edu>. Sat Nov 5 19:12:48 1994 Jason Merrill (jason@phydeaux.cygnus.com) * src/Makefile.in (LIBS): Remove. Tue Oct 18 17:51:14 1994 Per Bothner <bothner@kalessin.cygnus.com> * src/std-err.cc: Use stderror, instead of the non-standard sys_nerr and sys_errlist. Sat Sep 17 22:02:13 1994 Per Bothner (bothner@kalessin.cygnus.com) * src/key-list.cc (output_hash_function): Patch from William Bader <wbader@CSEE.Lehigh.Edu>. Fri Jul 15 09:38:11 1994 Per Bothner (bothner@cygnus.com) * src/std-err.cc: #include <errno.h>, and only declare extern int errno if errno is not a macro. Mon May 30 17:29:34 1994 Per Bothner (bothner@kalessin.cygnus.com) * Makefile.in (src_all, install): Make sure to add '/' after `pwd` in $rootme, as expected by FLAGS_TO_PASS. Wed May 11 00:47:22 1994 Jason Merrill (jason@deneb.cygnus.com) Make libg++ build with gcc -ansi -pedantic-errors * src/options.h: Lose commas at end of enumerator lists. Sun Dec 5 19:16:40 1993 Brendan Kehoe (brendan@lisa.cygnus.com) * src/hash-table.cc (Hash_Table::~Hash_Table): Don't pass an argument to fprintf, since it's not expecting one. Fri Nov 26 19:03:18 1993 Per Bothner (bothner@kalessin.cygnus.com) * src/list-node.cc: #undef index, for the sake of broken NeXT, Thu Nov 4 11:16:03 1993 Per Bothner (bothner@kalessin.cygnus.com) * Makefile.in (install): Use INSTALL_DATA for gperf.1. Mon Oct 25 18:40:51 1993 Per Bothner (bothner@kalessin.cygnus.com) * src/key-list.cc (Key_List::read_keys): Use POW macro to increase hash table size to power of 2. * options.h (LARGE_STACK_ARRAYS): New flag. Defaults to zero. * gen-perf.cc, key-list.cc, read-line.cc: Only stack-allocate large arrays if LARGE_STACK_ARRAYS is set. * main.cc (main): Only call setrlimit (RLIMIT_STACK, ...) if LARGE_STACK_ARRAYS. Mon Oct 4 17:45:08 1993 Per Bothner (bothner@kalessin.cygnus.com) * src/gen-perf.cc: Always use ANSI rand/srand instead of BSDisms. Wed Aug 18 12:19:53 1993 Per Bothner (bothner@kalessin.cygnus.com) * Makefile.in (src_all): Make less verbose output. Fri May 28 14:01:18 1993 Per Bothner (bothner@rtl.cygnus.com) * src/gen-perf.cc (Gen_Perf::change): Don't use gcc-specific 2-operand conditional expression. * src/key-list.cc (Key_List::output_lookup_array): Don't use variable-size stack arrays, unless compiled by g++. Tue May 4 14:08:44 1993 Per Bothner (bothner@cygnus.com) Changes (mostly from Peter Schauer) to permit compilation using cfront 3.0 and otherwise be ARM-conforming. * src/key-list.h: class Key_List must use public derivation of base class Std_Err (because Gen_Perf::operator() in gen-perf.cc calls Std_Err::report_error). * src/gen-perf.cc (Gen_Perf::affects_prev), src/hash-table.cc (Hash_Table::operator()): Don't use gcc-specific 2-operand conditional expression. * src/iterator.cc (Iterator::operator()): Don't use gcc-specific range construct in case label. * key-list.cc (Key_List::output_lookup_array, Key_List::read_keys), src/gen-perf.cc (Gen_Perf::operator(), src/read-line.cc (Read_Line::readln_aux): If not gcc, don't allocate variable-sized arrays on stack. * src/new.cc (operator new): Argument type should be size_t. * key-list.cc (Key_List::output_lookup_array, Key_List::read_keys), new/cc (::operator new): Don't use non-standard >?= operator. Tue Apr 27 20:11:30 1993 Per Bothner (bothner@cygnus.com) * src/Makefile.in: Define TARGETPROG, and use it. Mon Apr 19 00:29:18 1993 Per Bothner (bothner@cygnus.com) * Makefile.in, configure.in: Re-vamped configure scheme. * gperf.texinfo: Renamed to gperf.texi. * src/bool-array.{h,cc}: ANSIfy bzero->memset. Sat Jan 30 20:21:28 1993 Brendan Kehoe (brendan@lisa.cygnus.com) * tests/Makefile.in (mostlyclean): Also delete aout, cout, m3out, pout, and preout. Tue Dec 29 08:58:17 1992 Ian Lance Taylor (ian@cygnus.com) * Makefile.in: pass $(FLAGS_TO_PASS) to all calls to make. (FLAGS_TO_PASS): added INSTALL, INSTALL_DATA, INSTALL_PROGRAM. Mon Dec 21 18:46:46 1992 Per Bothner (bothner@rtl.cygnus.com) * tests/expected.* renamed to *.exp to fit in 14 chars. * tests/Makefile.in: Update accordingly. Also rename output.* to *.out. * src/Makefile.in (clean): Remove gperf program. Wed Dec 9 14:33:34 1992 Per Bothner (bothner@cygnus.com) * src/hash-table.cc, src/bool-array.h: ANSIfy bzero->memset. Thu Dec 3 19:34:12 1992 Per Bothner (bothner@cygnus.com) * Makefile.in (distclean, realclean): Don't delete Makefile before recursing. Fri Nov 6 13:41:49 1992 Per Bothner (bothner@rtl.cygnus.com) * key-list.{h,cc}: Remove MAX_INT (and similar) constant fields from Key_List class, and use INT_MAX (etc) from limits.h. * key-list.{h,cc}, options.{h,cc}, vectors.h: Removed all uses of initialized const fields, as they are non-standard - and their use was easy to do away with. Mostly, just made the constants static non-fields in the .cc file. Mon Nov 2 13:10:11 1992 Per Bothner (bothner@cygnus.com) * tests/Makefile.in: When generating cinset.c, don't pass -C, since -C assumes an ANSI compiler. Add the -C flag (with -a) when generating test.out.3 instead. * tests/expected.out.3: Update accordingly. Wed Aug 12 11:47:54 1992 Per Bothner (bothner@cygnus.com) * Makefile.in: Factor out common flags into $(FLAGS_TO_PASS). * Makefile.in: 'install-info' depends on gperf.info. Mon Aug 10 11:39:52 1992 Ian Lance Taylor (ian@dumbest.cygnus.com) * Makefile.in, src/Makefile.in: always create installation directories. Mon Jul 20 15:33:21 1992 Mike Stump (mrs@cygnus.com) * src/new.cc (operator new): Add cast from void * to char *, since it is not a standard conversion. Wed Jun 17 16:25:30 1992 Per Bothner (bothner@rtl.cygnus.com) * src/gen-perf.cc: #include <_G_config.h> for _G_SYSV. * src/key-list.cc: alloca() hair. * src/main.cc (main): Only call getrlimit if _G_HAVE_SYS_RESOURCE. * Makefile,in, {src,test}/Makefile.in: Fix *clean rules. Fri May 29 13:21:13 1992 Per Bothner (bothner@rtl.cygnus.com) * src/gen-perf.cc: Replace USG -> _G_SYSV. Thu May 14 13:58:36 1992 Per Bothner (bothner@rtl.cygnus.com) * src/Makefile.in: Don't pass obsolete flag -DUNLIMIT_STACK. * tests/Makefile.in (clean): Fix. Sat Mar 7 00:03:56 1992 K. Richard Pixley (rich@rtl.cygnus.com) * gperf.texinfo: added menu item hook. Wed Feb 26 18:04:40 1992 K. Richard Pixley (rich@cygnus.com) * Makefile.in, configure.in: removed traces of namesubdir, -subdirs, $(subdir), $(unsubdir), some rcs triggers. Forced copyrights to '92, changed some from Cygnus to FSF. Sun Jan 26 19:21:58 1992 Per Bothner (bothner at cygnus.com) * tests/Makefile.in: Use re-directed stdin instead of file name in argv. This allows us to remove the filename from the output, the expected output, and hence the diffs. (Note that the input file is in $(srcdir), which we cannot place in the expected out files.) * tests/expected.out.[1235]: Edit out input filename, to match new output. Thu Jun 28 16:17:27 1990 Doug Schmidt (schmidt at brilliant) * Wow, first fix on the new job! There was a dumb error in Key_List::output_lookup_function, where I printed the string "&wordlist[key]" instead of the correct "&wordlist[index]". * Added a couple of #ifdefs for USG support. Sun Jun 3 17:16:36 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Updated the version number to 2.5 and sent to Doug Lea for release with the latest GNU libg++. * Changed the error handling when a keyword file cannot be opened (now calls perror). Wed May 30 14:49:40 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Instrumented the source code with trace statements automagically inserted using my new automated trace instrumentation tool! Wed May 9 11:47:41 1990 Doug Schmidt (schmidt at siam.ics.uci.edu) * Really fixed the previous bug. Turns out that a small amount of logic had to be duplicated to handle static links that occur as part of dynamic link chains. What a pain!!! Tue May 8 23:11:44 1990 Doug Schmidt (schmidt at siam.ics.uci.edu) * Fixed a stupid bug in Key_List::output_lookup_array that was causing incorrect counts to be generated when there were both static and dynamic links occurring for the same hash value. Also simplified the code that performs the logic in this routine. Mon Apr 30 17:37:24 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Fixed stupid bug in Key_List::output_lookup_array that was making the generated lookup[] array contain `chars' even when the values stored in the chars are greater than 127! * Changed the behavior of the -G (global table) option so that it will output the `length[]' array in the global scope along with the `word_list[]' array. * Fixed a stupid bug in Key_List::output_lookup_function that would always output the complicated `duplicate-handling' lookup logic, even when there were no duplicates in the input! * Yikes, had to modify a bunch of stuff in key-list.cc to correctly handle duplicate entries. Changed the generated code so that the MIN_HASH_VALUE is no longer subtracted off when calculating the hash value for a keyword. This required changing some other code by substituting MAX_HASH_VALUE for TOTAL_KEYS in several places. Finally, this means that the generated tables may contain leading null entries, but I suppose it is better to trade-off space to get faster performance... Mon Mar 26 13:08:43 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Updated version number to 2.4 to reflect the latest changes. * Changed the main program so that it always prints out gperf's execution timings to the generated output file. Sun Mar 25 12:39:30 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Added the -Z option so that users can specify the name of the generated class explicitly. Updated documentation to reflect this change. * Modified the generated C++ class interface so that the functions are declared static (to remove the overhead of passing the `this' pointer). This means that operator()() can no longer be used, since it only works on non-static member functions. Also changed things so that there is no constructor (why waste the extra call, when it doesn't do anything, eh?) * Modified the behavior of Key_List::output when the -L C++ option is enabled. Previously the code generated use const data members to record MIN_WORD_LENGTH, MIN_HASH_VALUE, etc. However, as pointed out by James Clark this may result in suboptimal behavior on the part of C++ compilers that can't inline these values. Therefore, the new behavior is identical to what happens with -L C, i.e., either #defines or function-specific enums are used. Why sacrifice speed for some abstract notion of `code purity?' ;-) Tue Mar 6 18:17:42 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Added the -E option that defines constant values using an enum local to the lookup function rather than with #defines. This also means that different lookup functions can reside in the same file. Thanks to James Clark (jjc@ai.mit.edu). Sat Mar 3 20:19:00 1990 Doug Schmidt (schmidt at zola.ics.uci.edu) * Added a special case to key_list::output_switch that doesn't generate extra comparisons when the `-S' is given an argument of 1 (the normal case). This should speed up the generated code output a tad... Fri Feb 23 14:21:28 1990 Doug Schmidt (schmidt at zola.ics.uci.edu) * Renamed all instances of member function get_keysig_size to get_max_keysig_size, since this is more precise... * Changed all occurrences of charset to keysig (stands for ``key signature'') to reflect the new naming convention used in the USENIX paper. Thu Feb 22 11:28:36 1990 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Changed the name of the generated associated values table from asso_value to asso_values to reflect conventions in the USENIX C++ paper. Thu Feb 15 23:29:03 1990 Doug Schmidt (schmidt at zola.ics.uci.edu) * Updated the gperf.texinfo file to fix some formatting problems that had crept in since last time. Wed Feb 14 23:27:24 1990 Doug Schmidt (schmidt at zola.ics.uci.edu) * Fixed stupid bug in key-list.cc (get_special_input), wher gperf replaced each '%' with the succeeding character. * Added support for multiple target language generation. Currently handled languages are C and C++, with C as the default. Updated documentation and option handler to reflect the changes. * Added a global destructor to new.cc and removed the #ifdef, since the bloody thing now works with libg++. Mon Feb 14 13:00:00 1990 Doug Schmidt (schmidt at zola.ics.uci.edu) * Found out that my gperf paper was accepted at the upcoming USENIX C++ Conference in San Francisco. Yow! Tue Jan 30 09:00:29 1990 Doug Schmidt (schmidt at zola.ics.uci.edu) * #ifdef'd out the new.cc memory allocator, since there are problems with this and the libg++ stuff. * Changed key-list.h so that class Vectors is a public (rather than private) base class for class Key_List. The previous form was illegal C++, but wasn't being caught by the old g++ compiler. Should work now... ;-) Sun Dec 10 14:08:23 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added several changes from rfg@ics.uci.edu. These changes help to automate the build process. Wed Nov 15 15:49:33 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Removed conditional compilation for GATHER_STATISTICS. There's really no good reason to avoid collecting this info at run-time, since that section of code is *hardly* the bottleneck... ;-) * Simplified the C output routines in Key_List::set_output_types and Key_List::output_keyword_table a bit in order to speed-up and clean up the code generation. * Modified function Key_List::get_special_input so that it does not try to `delete' a buffer that turned out to be too short. This is important since the new memory management scheme does not handle deletions. However, adding a small amount of garbage won't hurt anything, since we generally don't do this operation more than a couple times *at most*! * Created a new file (new.cc) which includes my own overloaded operator new. This function should dramatically reduce the number of calls to malloc since it grabs large chunks and doles them out in small pieces. As a result of this change the class-specific `operator new' was removed from class List_Node. Tue Nov 14 21:45:30 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Continued to refine the great hack. The latest trick is to try and replace most uses of dynamic memory (i.e., calls to new) with uses of gcc dynamic arrays (i.e., an alloca solution). This makes life much easier for the overall process-size, since it reduces the amount of overhead for memory management. As a side-effect from this change there is no reason to have the Bool_Array::dispose member function, so it's outta here! * Fixed a stupid bug that was an disaster waiting to happen... Instead of making the boolean array large enough to index max_hash_value it was only large enough to index max_hash_value - 1. Once again, an off-by-one mistake in C/C++!!!! Mon Nov 13 19:38:27 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added the final great hack! This allows us to generate hash tables for near-perfect hash functions that contain duplicates, *without* having to use switch statements! Since many compilers die on large switch statements this feature is essential. Furthermore, it appears that the generated code is often *smaller* than that put out by compilers, even though a large, sparse array must be created. Here's the general idea: a. Generate the wordlist as a contiguous block of keywords, just as before when using a switch statement. This wordlist *must* be sorted by hash value. b. Generate the lookup array, which is an array of signed {chars,shorts,ints}, (which ever allows full coverage of the wordlist dimensions). If the value v, where v = lookup[hash(str,len)], is >= 0 and < TOTAL_KEYWORDS, then we simply use this result as a direct access into the wordlist array to snag the keyword for comparison. c. Otherwise, if v is < -TOTAL_KEYWORDS or > TOTAL_KEYWORDS this is an indication that we'll need to search through some number of duplicates hash values. Using a hash linking scheme we'd then index into a different part of the hash table that provides the starting index and total length of the duplicate entries to find via linear search! Sun Nov 12 13:48:10 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Simplified Key_List::output_min_max considerably by recognizing that since the keyword list was already sorted by hash value finding the min and max values is trivial! * Improved the debugging diagnostics considerably in classes Key_List, Hash_Table, and Gen_Perf. * Modified the `-s' option so that a negative argument is now interpreted to mean `allow the maximum associated value to be about x times *smaller* than the number of input keys.' This should help prevent massive explosion of generated hash table size for large keysets. Sat Nov 11 11:31:13 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added a field in class Key_List that counts the total number of duplicate keywords, both static and dynamic. * Added a new member function Bool_Array that deletes the dynamic memory allocated to Bool_Array::storage_array. This space may be needed for subsequent options, so it made sense to free it as soon as possible... * Renamed file/class Alpha_Vectors to Vectors, to avoid problems with 14 character length filenames on SYSV. Also changed file adapredefined.gperf to adadefs.gperf in the ./tests directory. * Modified class Options by changing the member function Options::total_positions to Options::get_charset_size and Options::set_charset_size. These two routines now either return the total charset size *or* the length of the largest keyword if the user specifies the -k'*' (ALLCHARS) option. This change cleans up client code. * Merged all the cperf changes into gperf. * Made sure to explicitly initialize perfect.fewest_collisions to 0. * Cleaned up some loose ends noticed by Nels Olson. 1. Removed `if (collisions <= perfect.fewest_collisions)' from Gen_Perf::affects_prev since it was superfluous. 2. Removed the fields best_char_value and best_asso_value from Gen_Perf. There were also unnecessary. 3. Fixed a braino in the Bool_Array::bool_array_reset function. Since iteration numbers can never be zero the `if (bool_array.iteration_number++ == 0)' must be `if (++bool_array.iteration_number == 0).' 4. Modified Std_Err::report_error so that it correctly handles "%%". * It is important to note that -D no longer enables -S. There is a good reason for this change, which will become manifested in the next release... (suspense!). * Made some subtle changes to Key_List::print_switch so that if finally seems to work correctly. Needs more stress testing, however... * Made a major change to the Key_List::print_switch function. The user can now specify the number of switch statements to generate via an argument to the -S option, i.e., -S1 means `generate 1 switch statement with all keywords in it,' -S2 means generate 2 switch statements with 1/2 the elements in each one, etc. Hopefully this will fix the problem with C compilers not being able to generate code for giant switch statements (but don't hold your breath!) * Changed Key_List::length function to Key_List::keyword_list_length. * Added a feature to main.c that prints out the starting wall-clock time before the program begins and prints out the ending wall-clock time when the program is finished. * Added the GATHER_STATISTICS code in hash-table.c so we can keep track of how well double hashing is doing. Eventually, GATHER_STATISTICS will be added so that all instrumentation code can be conditionally compiled in. * Fixed a stupid bug in Key_List::print_switch routine. This was necessary to make sure the generated switch statement worked correctly when *both* `natural,' i.e., static links and dynamic links, i.e., unresolved duplicates, hash to the same value. * Modified Bool_Array::~Bool_Array destructor so that it now frees the bool_array.storage_array when it is no longer needed. Since this array is generally very large it makes sense to return the memory to the freelist when it is no longer in use. * Changed the interface to constructor Hash_Table::Hash_Table. This constructor now passed a pointer to a power-of-two sized buffer that serve as storage for the hash table. Although this weakens information hiding a little bit it greatly reduces dynamic memory fragmentation, since we can now obtain the memory via a call to alloca, rather than malloc. This change modified Key_List::read_keys calling interface. * Since alloca is now being used more aggressively a conditional compilation section was added in main.c. Taken from GNU GCC, this code gets rid of any avoidable limit on stack size so that alloca does not fail. It is only used if the -DRLIMIT_STACK symbol is defined when gperf is compiled. * Added warnings in option.c so that user's would be informed that -r superceeds -i on the command-line. * Rewrote Gen_Perf::affects_prev. First, the code structure was cleaned up considerably (removing the need for a dreaded goto!). Secondly, a major change occurred so that Gen_Perf::affects_prev returns FALSE (success) when fewest_hits gets down to whatever it was after inserting the previous key (instead of waiting for it to reach 0). In other words, it stops trying if it can resolve the new collisions added by a key, even if there are still other old, unresolved collisions. This modification was suggested by Nels Olson and seems to *greatly* increase the speed of gperf for large keyfiles. Thanks Nels! * In a similar vein, inside the Gen_Perf::change routine the variable `perfect.fewest_collisions is no longer initialized with the length of the keyword list. Instead it starts out at 0 and is incremented by 1 every time change () is called. The rationale for this behavior is that there are times when a collision causes the number of duplicates (collisions) to increase by a large amount when it would presumably just have gone up by 1 if none of the asso_values were changed. That is, at the beginning of change(), you could initialize fewest_hits to 1+(previous value of fewest_hits) instead of to the number of keys. Thanks again, Nels. * Replaced alloca with new in the Gen_Perf::change function. This should eliminate some overhead at the expense of a little extra memory that is never reclaimed. * Renamed Gen_Perf::merge_sets to Gen_Perf::compute_disjoint_union to reflect the change in behavior. * Added the -e option so users can supply a string containing the characters used to separate keywords from their attributes. The default behavior is ",\n". * Removed the char *uniq_set field from LIST_NODE and modified uses of uniq_set in perfect.c and keylist.c. Due to changes to Gen_Perf::compute_disjoint_sets this field was no longer necessary, and its removal makes the program smaller and potentially faster. * Added lots of changes/fixes suggested by Nels Olson (umls.UUCP!olson@mis.ucsf.edu). In particular: 1. Changed Bool_Array so that it would dynamically create an array of unsigned shorts rather than ints if the LO_CAL symbol was defined during program compilation. This cuts the amount of dynamic memory usage in half, which is important for large keyfile input. 2. Added some additional debugging statements that print extra info to stderr when the -d option is enabled. 3. Fixed a really stupid bug in Key_List::print_switch A right paren was placed at the wrong location, which broke strlen (). 4. Fixed a subtle problem with printing case values when keylinks appear. The logic failed to account for the fact that there can be keylinks *and* regular node info also! 5. Changed the behavior of Key_List::read_keys so that it would honor -D unequivocally, i.e., it doesn't try to turn off dup handling if the user requests it, even if there are no immediate links in the keyfile input. 6. Modified the -j option so that -j 0 means `try random values when searching for a way to resolve collisions.' 7. Added a field `num_done' to the Gen_Perf struct. This is used to report information collected when trying to resolve hash collisions. 8. Modified the merge_sets algorithm to perform a disjoint union of two multisets. This ensures that subsequent processing in Gen_Perf::affect_prev doesn't waste time trying to change an associated value that is shared between two conflicting keywords. 9. Modified Gen_Perf::affects_prev so that it doesn't try random jump values unless the -j 0 option is enabled. 10. Fixed a silly bug in Gen_Perf::change. This problem caused gperf to seg fault when the -k* option was given and the keyfile file had long keywords. Sun Oct 29 00:18:55 1989 Doug Schmidt (schmidt at siam.ics.uci.edu) * Modified class-specific new operations for Read_Line and List_Node so they don't fail if SIZE is larger than twice the previous buffer size. Note we double buffer size everytime the previous buffer runs out, as a heuristic to reduce future calls to malloc. Sun Oct 22 13:49:43 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Updated gperf version number to 2.0. Send to Doug Lea for incorporation into the long-awaited `official' libg++ 1.36 release! * Thanks to Nels Olson a silly bug in Gen_Perf::change () was fixed. This problem caused gperf to seg fault when the -k* option was given and the keyfile file had long keywords. * Modified Key_List::print_hash_function so that it output max_hash_value + 1 (rather than just max_hash_value) for any associated value entries that don't correspond to keyword charset characters. This should speed up rejection of non-keyword strings a little in some cases. Sat Oct 21 19:28:36 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Fixed Key_List::print_hash_function so that it no longer output things like `return 0 + ...' Although this probably gets optimized away by even the worst C compilers there isn't any point tempting fate... ;-) * Fixed class List_Node's constructor so that it wouldn't a priori refuse to consider trying to hash keys whose length is less than the smallest user-specified key position. It turns out this is not a problem unless the user also specifies the -n (NOLENGTH) option, in which case such keys most likely don't have a prayer of being hashed correctly! * Changed the name of the generated lookup table from `Hash_Table' to `asso_value' to be consistent with the gperf paper. Tue Oct 17 14:19:48 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added a flag GATHER_STATISTICS in the Makefile. If defined during compilation this turns on certain collection facilities that track the performance of gperf during its execution. In particular, I want to see how many collisions occur for the double hashing Hash_Table. * Added a safety check so that we don't screw up if the total number of `resets' of the Bool_Array exceeds MAX_INT. Since this number is around 2^31 it is unlikely that this would ever occur for most input, but why take the risk? * Changed the behavior for the -a (ANSI) option so that the generated prototypes use int rather than size_t for the LEN parameter. It was too ugly having to #include <stddef.h> all over the place... Mon Oct 16 11:00:35 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Continued to work on the gperf paper for the USENIX C++ conference. At some point this will be merged back into the gperf documentation... Sat Oct 14 20:29:43 1989 Doug Schmidt (schmidt at siam.ics.uci.edu) * Added a majorly neat hack to Bool_Array, suggested by rfg. The basic idea was to throw away the Ullman array technique. The Ullman array was used to remove the need to reinitialize all the Bool_Array elements to zero everytime we needed to determine whether there were duplicate hash values in the keyword list. The current trick uses an `iteration number' scheme, which takes about 1/3 the space and reduces the overall program running a time by about 20 percent for large input! The hack works as follows: 1. Dynamically allocation 1 boolean array of size k. 2. Initialize the boolean array to zeros, and consider the first iteration to be iteration 1. 2. Then on all subsequent iterations we `reset' the bool array by kicking the iteration count by 1. 3. When it comes time to check whether a hash value is currently in the boolean array we simply check its index location. If the value stored there is *not* equal to the current iteration number then the item is clearly *not* in the set. In that case we assign the iteration number to that array's index location for future reference. Otherwise, if the item at the index location *is* equal to the iteration number we've found a duplicate. No muss, no fuss! Mon Oct 2 12:30:54 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Changed some consts in options.h to enumerals, since g++ doesn't seem to like them at the moment! Sat Sep 30 12:55:24 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Fixed a stupid bug in Key_List::print_hash_function that manifested itself if the `-k$' option was given (i.e., only use the key[length] character in the hash function). * Added support for the -C option. This makes the contents of all generated tables `readonly'. * Changed the handling of generated switches so that there is only one call to str[n]?cmp. This *greatly* reduces the size of the generated assembly code on all compilers I've seen. * Fixed a subtle bug that occurred when the -l and -S option was given. Code produced looked something like: if (len != key_len || !strcmp (s1, resword->name)) return resword; which doesn't make any sense. Clearly, this should be: if (len == key_len && !strcmp (s1, resword->name)) return resword; Tue Sep 26 10:36:50 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Changed class Read_Line's definition so that it no longer needs to know about the buffering scheme used to speed up dynamic memory allocation of input keywords and their associated attributes. This means that operator new is no longer a friend of Read_Line. Mon Sep 25 23:17:10 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Decided that Obstacks had too much overhead, so they were removed in favor of super-efficient, low-overhead buffered storage allocation hacks in Read_Line and List_Node. * No longer try to inline functions that g++ complains about (Key_List::Merge and Key_List::Merge_Sort). Sun Sep 24 13:11:24 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Changed classes Read_Line and List_Node to use Obstacks in order to cache memory allocation for keyword strings and List_Nodes. * Continued to experiment with inheritance schemes. * Added a new file `alpha.h', that declares static data shared (i.e., inherited) between classes List_Node and Key_List. Tue Sep 12 16:14:41 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Made numerous changes to incorporate multiple inheritance in gperf. Wed Aug 16 23:04:08 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added the -DCOMPILER_FIXED flag to the ./src/Makefile. This implies that people trying to compile gperf need to have a working version of the new g++ compiler (1.36.0). * Removed some extra spaces that were being added in the generated C code. Mon Jul 24 17:09:46 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Fixed PRINT_HASH_FUNCTION and PRINT_LOOKUP_FUNCTION in keylist.c so that the generated functions take an unsigned int length argument. If -a is enabled the prototype is (const char *str, size_t len). Fri Jul 21 13:06:15 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Fixed a typo in PRINT_KEYWORD_TABLE in keylist.cc that prevented the indentation from working correctly. * Fixed a horrible typo in PRINT_KEYWORD_TABLE in keylist.cc that prevented links from being printed correctly. Tue Jul 18 16:04:31 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Fixed up readline.cc and readline.h so that they work OK with g++ compilers that aren't completely up-to-date. If symbol COMPILER_FIXED is defined then the behavior that works on my more recent version of g++ is enabled. Sun Jul 9 17:53:28 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Changed the ./tests subdirectory Makefile so that it uses $(CC) instead of gcc. Sun Jul 2 21:52:15 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Fixed a number of subtle bugs that occurred when -S was combined with various and sundry options. * Added the -G option, that makes the generated keyword table a global static variable, rather than hiding it inside the lookup function. This allows other functions to directly access the contents in this table. * Added the "#" feature, that allows comments inside the keyword list from the input file. Comment handling takes place in readline.c. This simplifies the code and reduces the number of malloc calls. * Also added the -H option (user can give the name of the hash function) and the -T option (prevents the transfer of the type decl to the output file, which is useful if the type is already defined elsewhere). Thu Jun 22 20:39:39 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Modified many classes so that they would inherit Std_Err as a base class. This makes things more abstract... Fri Jun 16 14:23:00 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Modified the -f (FAST) option. This now takes an argument. The argument corresponds to the number of iterations used to resolve collisions. -f 0 uses the length of the keyword list (which is what -f did before). This makes life much easier when dealing with large keyword files. Tue Jun 6 17:53:27 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added the -c (comparison) option. Enabling this will use the strncmp function for string comparisons. The default is to use strcmp. * Fixed a typo in key_list.cc (PRINT_SWITCH). This caused faulty C code to be generated when the -D, -p, and -t options were all enabled. Thu May 25 14:07:21 1989 Doug Schmidt (schmidt at siam.ics.uci.edu) * Once again, changed class Read_Line to overload global operator new. Hopefully, this will work...! Sun May 21 01:51:45 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Modified Key_List::print_hash_function () so that it properly formats the associated values in the hash table according to the maximum number of digits required to represent the largest value. * Removed the named return value from class Hash_Table's operator (), since this causes a seg fault when -O is enabled. No sense tripping subtle g++ bugs if we don't have to.... ;-) * Removed the operator new hack from Read_Line, since this seemed to create horrible bus error problems. * Changed many class member functions and data members to be `static', if they don't manipulate this! Fri May 12 23:06:56 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Changed class Std_Err to use static member functions, a la Ada or Modula 2. This eliminates the need for an explicit error-handler class object. * Added the ``named return value'' feature to Hash_Table::operator () and Bool_Array::operator [], just for the heck of it.... ;-) * Changed the previous hack in Read_Line so that we now use the overloaded global `new' instead of NEW_STRING! Wed May 3 17:36:55 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Updated to version 1.7. This reflects the recent major changes and the new C port. * Modified the GNU getopt.cc routine to have a class-based interface. * Fixed a typo in Perfect.cc ~Perfect that prevented the actual maximum hash table size from being printed (maybe the stream classes weren't so bad after all.... ;-). * Added support for the -f option. This generates the perfect hash function ``fast.'' It reduces the execution time of gperf, at the cost of minimizing the range of hash values. Tue May 2 16:23:29 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Added an efficiency hack to Read_Line. Instead of making a call to operator NEW (a.k.a. malloc) for each input string a new member function NEW_STRING stores a large buffer from which new strings are carved out, growing the buffer if necessary. It might be useful to add this throughout the program.... * Removed all unnecessary calls to DELETE. If the program is about to exit it is silly to waste time freeing memory. * Added the GNU getopt program to the distribution. This makes GPERF portable to systems that don't include getopt in libc. * Added a strcspn member to class Key_List. This also increases portability. * Added the get_include_src function from keylist.c as a member function in class Key_List. Hopefully every function is now associated with a class. This aids abstraction and modularity. * Ported gperf to C. From now on both K&R C and GNU G++ versions will be supported. There will be two ChangeLog files, one for each version of the program. Mon May 1 16:41:45 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Fixed a bug with -k'*'. This now prints out *all* the cases up to the length of the longest word in the keyword set. Sun Apr 30 12:15:25 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Removed all use of the stream classes. Too ugly, slow, and not handled by the c++-mode formatter.... * Modified the handling of links (i.e., keywords that have identical hash values as other keywords). This should speed up hash function generation for keyword sets with many duplicate entries. The trick is to treat duplicate values as equivalence classes, so that each set of duplicate values is represented only once in the main list processing. * Fixed some capitialization typos and indentations mistakes in Key_List::print_hash_function. Sat Apr 29 12:04:03 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Fixed a typo/logico in Key_List::print_switch that prevented the last keyword in the keyword list to be print out. This requires further examination..... * Fixed a stupid bug in List_Node::List_node. If the -k'*' option was enabled the KEY_SET string wasn't getting terminated with '\0'! Fri Apr 28 12:38:35 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Renamed strexp.h and strexp.cc to iterator.h and iterator.cc. Also changed the strexp class to iterator. Continued to work on style... * Updated the version number to 1.6. This reflects all the recent changes. Thu Apr 27 00:14:51 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added the -D option that properly handles keyword sets that contain duplicate hash values. * Continued the stylistic changes. Added the #pragma once directive to all the *.h files. Removed all #defines and replaced them with static consts. Also moved the key_sort routine from options.cc into the options class as a member function. Mon Apr 3 13:26:55 1989 Doug Schmidt (schmidt at zola.ics.uci.edu) * Made massive stylistic changes to bring source code into conformance with GNU style guidelines. Thu Mar 30 23:28:45 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Fixed up the output routines so that they generate code corresponding to the GNU style guidelines. Sat Mar 11 13:12:37 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Fixed Stderr constructors so that they wouldn't try to use the base class initializer syntax for the static class variable Program_Name. G++ 1.34 is stricter in enforcing the rules! Fri Mar 10 11:24:14 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Removed -v and ``| more'' from the Makefile to keep rfg happy... Thu Mar 2 12:37:30 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Sent latest GNU gperf version 1.5 to Doug Lea for inclusion into libg++ 1.34. Note that there is a small bug with the new %{ ... %} source inclusion facility, since it doesn't understand comments and will barf if %{ or %} appear nested inside the outermost delimiters. This is too trivial of a defect to fix at the moment... Tue Feb 28 11:19:58 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added the -K option, which allows the user to provide a alternative name for the keyword structure component. The default is still ``name.'' * Added the LEX and YACC-like ability to include arbitrary text at the beginning of the generated C source code output. This required two new functions Get_Special_Input, Key_List::Save_Include_Src; * Fixed memory allocation bug in Key_List::Set_Types. Variable Return_Type needs 1 additional location to store the "*" if the -p option is used. * Added code to NULL terminate both Struct_Tag and Return_Type, *after* the strncpy (stupid mistake). Mon Feb 27 14:39:51 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added a new option -N. This allows the user to specify the name to be used for the generated lookup function. The default name is still ``in_word_set.'' This makes it possible to completely automate the perfect hash function generation process! Mon Feb 20 23:33:14 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Corrected the Hash_Table::operator () function so that *it* is responsible for deciding when a new key has the same signature as a previously seen key. The key length information is now used internally to this function to decide whether to add to the hash table those keys with the same key sets, but different lengths. Before, this was handled by the Key_List::Read_Keys function. However, this failed to work for certain duplicate keys, since they weren't being entered into the hash table properly. Sun Feb 19 16:02:51 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Modified class Options by moving the enum Option_Type out of the class. This is to satisfy the new enumeration scope rules in C++. Sun Jan 15 15:12:09 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Incremented the version number upto 1.4 to reflect the new options that affect the generated code. Send the new distribution off to Michael for use with g++ 1.33. * Added a fix to Key_List::Read_Keys so that it checks for links properly when the -n option is used. Previously, it didn't catch obvious links, which caused it to spend large amount of time searching for a solution that could never occur! * Modified the Key_List data structure to record *both* the minimum and the maximum key lengths. This information is now computed in Key_List::Read_Keys, and thus Key_List::Print_Min_Max doesn't need to bother. * Modifed the key position iterator scheme in options.cc to eliminate the need for member function Options::Advance. Now, the Options::Get function performs the advancement automatically, obviating the need for an extra function call. * Added the new function Options::Print_Options, to print out the user-specified command line options to generated C output file. * Added a new function, Key_List::Print_Keylength_Table, which creates a table of lengths for use in speeding up the keyword search. This also meant that a new option, -l (LENTABLE) is recognized. It controls whether the length table is printed and the comparison made in the generated function ``in_word_set.'' * Added a comment at the top of the generated C code output file that tells what version of gperf was used. Next, I'll also dump out the command line options as a comment too. Thanks to Michael Tiemann for the feedback on this. * Fixed the -n option to make it work correctly with other parts of the program (most notably the Perfect::Hash function and the computation of minimum and maximum lengths. Fri Jan 13 21:25:27 1989 Doug Schmidt (schmidt at siam.ics.uci.edu) * Realized the the need to add a test that will enable optimziation of the generated C code in the ``hash'' function by checking whether all the requested key positions are guaranteed to exist due to the comparison in `in_word_set.'' I'll put this in soon.... Thu Jan 12 20:09:21 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Added pascal, modula3, and modula2 tests inputs to the Makefile * Recognised that there is a bug with the -n option. However I'm too busy to fix it properly, right now. The problem is that the generated #define end up being 0, since that's my hack to make -n work. This needs complete rethinking! Tue Jan 10 00:08:16 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Added a new option, -n, that instructs gperf to not use the length of an identifier when computing the hash functions. I'm not sure how useful this is! * Retransmitted the distribution to rocky.oswego.edu. Hopefully, this will work! * Began fixing the indentation and capitalization to conform to the GNU coding guidelines. Mon Jan 9 22:23:18 1989 Doug Schmidt (schmidt at pompe.ics.uci.edu) * Fixed horrible bug in Read_Line::Readln_Aux. This was a subtle and pernicous off-by-1 error, that overwrote past the last character of the input string buffer. I think this fault was killing the vax! * Yow, fixed an oversight in List_Node::List_Node, where the pointer field Next was uninitialized. Luckily, the new routine seems to return 0 filled objects the first time through! Sun Jan 8 13:43:14 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Modified the ``key linked'' diagnostic in Key_List::Read_Keys to be more helpful and easy to read. * Fixed the List_Node::List_Node so that it would ignore trailing fields if the -t option was not enabled. * Moved the List_Node declarations out of keylist.h and into a file of its own, called listnode.cc and listnode.h Made Set_Sort a member function of class List_Node. * Massively updated the documentation in the gperf.texinfo file. * Polished off the major revision to the print functions, added a few new tests in the Makefile to check for the validity of the program and ftp'ed the entire distribution off to Doug Lea for libg++. ( changed it to 1.3 to reflect the major changes with the generated C code ). * Fixed Key_List::Print_Switch to deal with the -p and -t options. This meant that the ``still-born'' function Key_List:: Print_Type_Switch was superflous, so I removed it. Also, removed the restriction in Option that the -p and -t options couldn't be used simultaneously. * Modified List_Node::List_Node, to perform only 1 call to ``new'' when dynamically allocating memory for the Key_Set and the Uniq_Set. Sat Jan 7 14:10:51 1989 Doug Schmidt (schmidt at glacier.ics.uci.edu) * Fixed a big bug with the new policy of nesting the wordlist inside of generated function ``in_word_set.'' I'd forgotten to declare the wordlist array as static! ( arrgh ). * Added a new function Key_List::Set_Types, that figures out the return type for generated function ``in_word_set,'' the user-defined ``struct tag,'' if one is used, and also formates the array type for the static local array. * Changed the print routines to take advantage of the new -p option. * Began adding the hooks to allow the return of a pointer to a user defined struct location from the generated ``in_word_set'' function instead of the current 0 or 1 return value. Created function Key_List::Print_Type_Switch and added option -p to class Option, allowing the user to request generation of the aforementioned pointers returned instead of booleans. * Put in checks in class Option to make sure that -S and -t options are not used simultaneously. This restriction will be removed in subsequent releases, once I decide on a clean way to implement it. * Sent version 1.2 to Doug Lea for possible inclusion into the libg++ distribution. * Moved the static word_list array inside the generated function in_word_set. This supports better data hiding. * Added a texinfo file, gperf.texinfo * Revised the Makefile to cleanup the droppings from texinfo and changed the name of gperf.cc and gperf.h to perfect.cc and perfect.h. Fri Jan 6 13:04:45 1989 Doug Schmidt (schmidt at crimee.ics.uci.edu) * Implemented the switch statement output format. Much better for large datasets in terms of space used. * Added new functions to break up the Key_List::Output function. Functions added were Key_List::Print_Switch, Key_List::Print_Min_Max, Key_List::Print_Keyword_Table, Key_List::Print_Hash_Function, and Key_List::Print_Lookup_Function. This simplifies the big mess in Key_List::Output considerably! * Added switch statement option to Options, which potentially trades time for space in the generated lookup code. Thu Jan 5 22:46:34 1989 Doug Schmidt (schmidt at siam.ics.uci.edu) * Released version 1.1 * Fixed a bug with Gperf::Merge_Set, it was skipping letters shared between the Set_1 and Set_2. * Added the optimal min/max algorithm in Key_List::Output. This runs in O ( 3n/2 ), rather than O ( 2n ) time. * Changed Gperf::Sort_Set to use insertion sort, rather than bubble sort. * Added a check in Key_List::Output for the special case where the keys used are 1,$. It is possible to generate more efficient C code in this case.