COIN-OR::LEMON - Graph Library

source: glpk-cmake/NEWS

Last change on this file was 1:c445c931472f, checked in by Alpar Juttner <alpar@…>, 12 years ago

Import glpk-4.45

  • Generated files and doc/notes are removed
File size: 68.0 KB
1GLPK 4.45 (release date: Dec 05, 2010)
3        This is a bug-fix release.
5        Several bugs/typos were fixed. Thanks to
6        Xypron <>,
7        Robbie Morrison <>, and
8        Ali Baharev <> for reports.
10        Some glpk documents was re-formatted and merged into a single
11        document. Now the glpk documentation consists of the following
12        three main documents (all included in the distribution):
14        GLPK: Reference Manual
16        GLPK: Graph and Network Routines
18        Modeling Language GNU MathProg: Language Reference
20GLPK 4.44 (release date: Jun 03, 2010)
22        The following suffixes for variables and constraints were
23        implemented in the MathProg language:
25        .lb     (lower bound),
26        .ub     (upper bound),
27        .status (status in the solution),
28        .val    (primal value), and
29        .dual   (dual value).
31        Thanks to Xypron <> for draft implementation
32        and testing.
34        Now the MathProg language allows comment records (marked by
35        '#' in the very first position) in CSV data files read with the
36        table statements. Note that the comment records may appear only
37        in the beginning of a CSV data file.
39        The API routine glp_cpp to solve the Critical Path Problem was
40        added and documented.
42GLPK 4.43 (release date: Feb 20, 2010)
44        This is a maintainer release.
46        `' was changed to allow building the package under
47        Mac OS and Darwin with ODBC support.
48        Thanks to Xypron <> for suggestions and Noli
49        Sicad <> for testing.
51        The SQL table driver was improved to process NULL data. Thanks
52        to Xypron <>.
54        Some bugs were fixed in the LP/MIP preprocessor.
56GLPK 4.42 (release date: Jan 13, 2010)
58        The following new API routines were added:
60        glp_check_dup         check for duplicate elements in sparse
61                              matrix
62        glp_sort_matrix       sort elements of the constraint matrix
63        glp_read_prob         read problem data in GLPK format
64        glp_write_prob        write problem data in GLPK format
65        glp_analyze_bound     analyze active bound of non-basic variable
66        glp_analyze_coef      analyze objective coefficient at basic
67                              variable
68        glp_print_ranges      print sensitivity analysis report (this
69                              routine replaces lpx_print_sens_bnds and
70                              makes it deprecated)
72        For description of these new routines and the GLPK LP/MIP
73        format see a new edition of the reference manual included in
74        the distribution. (Chapter "Graph and network API routines" was
75        carried out from the main reference manual and included in the
76        distribution as a separate document.)
78        The following new command-line options were added to the stand-
79        alone solver glpsol:
80        --glp filename        read problem data in GLPK format
81        --wglp filename       write problem data in GLPK format
82        --ranges filename     print sensitivity analysis report (this
83                              option replaces --bounds)
85        Now all GLPK routines performing file I/O support special
86        filenames "/dev/stdin", "/dev/stdout", and "/dev/stderr", which
87        can be specified in the same way as regular filenames. This
88        feature is plaform-independent.
90GLPK 4.41 (release date: Dec 21, 2009)
92        The following new API routies were added:
94        glp_transform_row     transform explicitly specified row
95        glp_transform_col     transform explicitly specified column
96        glp_prim_rtest        perform primal ratio test
97        glp_dual_rtest        perform dual ratio test
99        For description of these new routines see a new edition of the
100        reference manual included in the distribution.
102        The following API routines are deprecated: lpx_transform_row,
103        lpx_transform_col, lpx_prim_ratio_test, lpx_dual_ratio_test.
105        Some improvements were made in the MIP solver (glp_intopt).
107        The SQL table driver used to read/write data in MathProg models
108        was changed to allow multiple arguments separated by semicolon
109        in SQL statements. Thanks to Xypron <>.
111        Two new options were added to the glpsol stand-alone solver:
112        --seed value (to initialize the pseudo-random number generator
113        used in MathProg models with specified value), and
114        --ini filename (to use a basis previously saved with -w option
115        as an initial basis on solving similar LP's).
117        Two new MathProg example models were included. Thanks to
118        Nigel Galloway <> and Noli Sicad
119        <> for contribution.
121        Scripts to build GLPK with Microsoft Visual Studio 2010 for
122        both 32-bit and 64-bit Windows were included. Thanks to Xypron
123        <> for contribution and testing.
125GLPK 4.40 (release date: Nov 03, 2009)
127        The following new API routines were added:
129        glp_del_vertices      remove vertices from graph
130        glp_del_arc           remove arc from graph
131        glp_wclique_exact     find maximum weight clique with the exact
132                              algorithm developed by Prof. P. Ostergard
133        glp_read_ccdata       read graph in DIMACS clique/coloring
134                              format
135        glp_write_ccdata      write graph in DIMACS clique/coloring
136                              format
138        For description of these new routines see a new edition of the
139        reference manual included in the distribution.
141        The hybrid pseudocost branching heuristic was included in the
142        MIP solver. It is available on API level (iocp.br_tech should
143        be set to GLP_BR_PCH) and in the stand-alone solver glpsol
144        (via the command-line option --pcost). This heuristic may be
145        useful on solving hard MIP instances.
147        The branching heuristic by Driebeck and Tomlin (used in the
148        MIP solver by default) was changed to switch to branching on
149        most fractional variable if an lower bound of degradation of
150        the objective is close to zero for all branching candidates.
152        A bug was fixed in the LP preprocessor (routine npp_empty_col).
153        Thanks to Stefan Vigerske <> for the
154        bug report.
156        A bug was fixed and some improvements were made in the FPUMP
157        heuristic module. Thanks to Xypron <>.
159        A bug was fixed in the API routine glp_warm_up (dual
160        feasibility test was incorrect in maximization case). Thanks to
161        Uday Venkatadri <> for the bug report.
163GLPK 4.39 (release date: Jul 26, 2009)
165        The following new API routines were added:
167        glp_warm_up           "warm up" LP basis
168        glp_set_vertex_name   assign (change) vertex name
169        glp_create_v_index    create vertex name index
170        glp_find_vertex       find vertex by its name
171        glp_delete_v_index    delete vertex name index
172        glp_read_asnprob      read assignment problem data in DIMACS
173                              format
174        glp_write_asnprob     write assignment problem data in DIMACS
175                              format
176        glp_check_asnprob     check correctness of assignment problem
177                              data
178        glp_asnprob_lp        convert assignment problem to LP
179        glp_asnprob_okalg     solve assignment problem with the
180                              out-of-kilter algorithm
181        glp_asnprob_hall      find bipartite matching of maxumum
182                              cardinality with Hall's algorithm
184        Also were added some API routines to read plain data files.
186        The API routines glp_read_lp and glp_write_lp to read/write
187        files in CPLEX LP format were re-implemented. Now glp_write_lp
188        correctly writes double-bounded (ranged) rows by introducing
189        slack variables rather than by duplicating the rows.
191        For description of these new routines see a new edition of the
192        reference manual included in the distribution.
194        The 'xfree(NULL)' bug was fixed in the AMD routines. Thanks to
195        Niels Klitgord <> for bug report.
197        The message "Crashing..." was changed to "Constructing initial
198        basis..." due to suggestion by Thomas Kahle <>.
200        Some typos were corrected in glpsol output messages. Thanks to
201        Xypron <> for patch.
203GLPK 4.38 (release date: May 02, 2009)
205        API routines glp_read_mps and glp_write_mps were improved.
207        Some improvements were made in the dual simplex routines.
209        Two external software modules AMD and COLAMD were included in
210        the distribution (for more details please see src/amd/README
211        and src/colamd/README). Now they are used in the interior-point
212        solver to reorder the matrix prior to Cholesky factorization.
214        API routine glp_ipt_status may return two new statuses due to
215        changes in the routine glp_interior. For details please see the
216        reference manual included in the distribution.
218        A minor bug was fixed in the graph/network routines. Thanks to
219        Nelson H. F. Beebe <> for bug report.
221GLPK 4.37 (release date: Mar 29, 2009)
223        The 0-1 Feasibility Pump heuristic was included in the GLPK
224        integer optimizer glp_intopt. On API level the heuristic can be
225        enabled by setting the parameter fp_heur in glp_iocp to GLP_ON.
226        This feature is also available in the solver glpsol through
227        command-line option '--fpump'. For more details please see the
228        reference manual included in the distribution.
230        The following new API routines were added:
232        glp_print_sol         write basic solution in printable format
233        glp_print_ipt         write interior-point solution in printable
234                              format
235        glp_print_mip         write MIP solution in printable format
236        glp_read_graph        read (di)graph from plain text file
237        glp_write_graph       write (di)graph to plain text file
238        glp_weak_comp         find all weakly connected components
239        glp_strong_comp       find all strongly connected components
241        The following API routines are deprecated: lpx_print_sol,
242        lpx_print_ips, lpx_print_mip, lpx_print_prob (the latter is
243        equivalent to glp_write_lp).
245        A bug was fixed in the interior-point solver (glp_interior) to
246        correctly compute dual solution components when the problem is
247        scaled.
249        The files and were changed:
250        (a) to allow using autoreconf/autoheader;
251        (b) to allow building the package in a directory other than its
252            source directory.
253        Thanks to Marco Atzeri <> for bug report.
255        An example model in the GNU MathProg language was added.
256        Thanks to Larry D'Agostino <Larry.D'> for
257        contribution.
259GLPK 4.36 (release date: Feb 06, 2009)
261        The following new API routines were added to the package:
263        glp_mincost_okalg     find minimum-cost flow with out-of-kilter
264                              algorithm
265        glp_maxflow_ffalg     find maximal flow with Ford-Fulkerson
266                              algorithm
268        For detailed description of these new routines and related data
269        structures see chapter "Graph and Network API Routines" in a new
270        edition of the reference manual included in the distribution.
272        The following two new command-line options were added to the
273        solver glpsol:
275        --mincost             read min-cost flow data in DIMACS format
276        --maxflow             read maximum flow data in DIMACS format
278        Duplicate symbols in the header glpk.h were removed to allow
279        using swig.
280        Thanks to Kelly Westbrooks <> and
281        Nigel Galloway <> for suggestion.
283        A minor defect was fixed in the routine glp_write_lp.
284        Thanks to Sebastien Briais <> for bug report.
286        A minor bug was fixed in the SQL module.
287        Thanks to Xypron <> for patch.
289        Some new example models in the GNU MathProg modeling language
290        were added. Thanks to Sebastian Nowozin <> and
291        Nigel Galloway <> for contribution.
293GLPK 4.35 (release date: Jan 09, 2009)
295        The following new API routines were added to the package:
297        glp_create_graph      create graph
298        glp_set_graph_name    assign (change) graph name
299        glp_add_vertices      add new vertices to graph
300        glp_add_arc           add new arc to graph
301        glp_erase_graph       erase graph content
302        glp_delete_graph      delete graph
303        glp_read_mincost      read minimum cost flow problem data in
304                              DIMACS format
305        glp_write_mincost     write minimum cost flow problem data in
306                              DIMACS format
307        glp_mincost_lp        convert minimum cost flow problem to LP
308        glp_netgen            Klingman's network problem generator
309        glp_gridgen           grid-like network problem generator
310        glp_read_maxflow      read maximum flow problem data in DIMACS
311                              format
312        glp_write_maxflow     write maximum flow problem data in DIMACS
313                              format
314        glp_maxflow_lp        convert maximum flow problem to LP
315        glp_rmfgen            Goldfarb's maximum flow problem generator
317        For detailed description of these new routines and related data
318        structures see chapter "Graph and Network API Routines" in a new
319        edition of the reference manual included in the distribution.
321        A minor change were made in the internal routine xputc. Thanks
322        to Luiz Bettoni <> for suggestion.
324        A minor bug was fixed in the internal routine mpl_fn_time2str.
325        Thanks to Stefan Vigerske <> for bug report.
327GLPK 4.34 (release date: Dec 04, 2008)
329        The GNU MathProg modeling language was supplemented with three
330        new built-in functions:
332        gmtime    obtaining current calendar time
333        str2time  converting character string to calendar time
334        time2str  converting calendar time to character string
336        (Thanks to Xypron <>.)
338        For detailed description of these functions see Appendix A in
339        the document "Modeling Language GNU MathProg", a new edition of
340        which was included in the distribution.
342        A bug was fixed in the MIP solver. Thanks to Nigel Galloway
343        <> for bug report.
345        A new makefile was added to build the GLPK DLL with Microsoft
346        Visual Studio Express 2008 for 64-bit Windows. Thanks to Xypron
347        <> for contribution and testing.
349GLPK 4.33 (release date: Oct 30, 2008)
351        The following new API routines were added to the package:
352        glp_copy_prob         copy problem object content
353        glp_exact             solve LP in exact arithmetic
354                              (makes lpx_exact deprecated)
355        glp_get_unbnd_ray     determine variable causing unboundedness
356                              (makes lpx_get_ray_info deprecated)
357        glp_interior          solve LP with interior-point method
358                              (makes lpx_interior deprecated)
360        The following new API routines for processing models written in
361        the GNU Mathprog language were added to the package:
362        glp_mpl_alloc_wksp    allocate the translator workspace
363        glp_mpl_read_model    read and translate model section
364        glp_mpl_read_data     read and translate data section
365        glp_mpl_generate      generate the model
366        glp_mpl_build_prob    build LP/MIP instance from the model
367        glp_mpl_postsolve     postsolve the model
368        glp_mpl_free_wksp     deallocate the translator workspace
369        (These routines make lpx_read_model deprecated.)
371        For description of all these new API routines see the reference
372        manual included in the distribution.
374        A crude implementation of CPLEX-like interface to GLPK API was
375        added to the package. Currently it allows using GLPK as a core
376        LP solver for Concorde, a well known computer code for solving
377        the symmetric TSP. For details see examples/cplex/README.
379        Some bugs were fixed in the SQL table driver. Thanks to Xypron
380        <>.
382GLPK 4.32 (release date: Oct 03, 2008)
384        The following new features were included in the MIP solver
385        (the API routine glp_intopt):
387        *  MIP presolver
388        *  mixed cover cut generator
389        *  clique cut generator
390        *  Euclidean reduction of the objective value
392        Due to changes the routine glp_intopt may additionally return
395        The API routines lpx_integer are lpx_intopt are deprecated,
396        since they are completely superseded by glp_intopt.
398        The following new branch-and-cut API routines were added:
399        glp_ios_row_attr      determine additional row attributes
400        glp_ios_pool_size     determine current size of the cut pool
401        glp_ios_add_row       add constraint to the cut pool
402        glp_ios_del_row       delete constraint from the cut pool
403        glp_ios_clear_pool    delete all constraints from the cut pool
405        For description of these new routines see the reference manual
406        included in the distribution.
408        The stand-alone solver glpsol was changed to allow multiple
409        data files.
411        A new edition of the supplement "Using Data Tables in the GNU
412        MathProg Modeling Language" was included.
414        As usual, some bugs were fixed (in the MathProg translator).
415        Thanks to Xypron <>.
417GLPK 4.31 (release date: Sep 02, 2008)
419        The core LP solver based on the dual simplex method was
420        re-implemented and now it provides both phases I and II.
422        The following new API routines were added:
423        glp_scale_prob  automatic scaling of problem data
424        glp_std_basis   construct standard initial LP basis
425        glp_adv_basis   construct advanced initial LP basis
426        glp_cpx_basis   construct Bixby's initial LP basis
428        For description of these new routines see the reference manual
429        included in the distribution.
431        The following API routines are deprecated:
432        lpx_scale_prob, lpx_std_basis, lpx_adv_basis, lpx_cpx_basis.
434        Necessary changes were made in memory allocation routines to
435        resolve portability issues for 64-bit platforms.
437        New version of the routine lpx_write_pb to write problem data
438        in OPB (pseudo boolean format) was added to the package. Thanks
439        to Oscar Gustafsson <> for the contribution.
441        Two new makefiles were added to build the package for 32- and
442        64-bit Windows with Microsoft Visual Studio Express 2008.
443        Thanks to Heinrich Schuchardt <> (aka
444        Xypron) for the contribution and testing.
446        Two new makefiles were added to build the package with Digital
447        Mars C/C++ 8.50 and Open Watcom C/C++ 1.6 (for 32-bit Windows).
449GLPK 4.30 (release date: Aug 13, 2008)
451        The core LP solver based on the primal simplex method was
452        re-implemented to allow its further improvements. Currently the
453        new version provides the same features as the old one, however,
454        it is a bit faster and more numerically stable.
456        Some changes were made in the MathProg translator to allow <,
457        <=, >=, and > on comparing symbolic values. Thanks to Heinrich
458        Schuchardt <> for patches.
460        Internal routine set_d_eps in the exact LP solver was changed
461        to prevent approximation errors in case of integral data.
462        Thanks to Markus Pilz <> for bug report.
464GLPK 4.29 (release date: Jul 06, 2008)
466        The configure script was changed to disable all optional
467        features by default. For details please see file INSTALL.
469        The following new API routines were added:
470        glp_erase_prob  erase problem object content
471        glp_read_mps    read problem data in MPS format
472        glp_write_mps   write problem data in MPS format
473        glp_read_lp     read problem data in CPLEX LP format
474        glp_write_lp    write problem data in CPLEX LP format
476        For description of these new routines see the reference manual
477        included in the distribution.
479        The following API routines are deprecated:
480        lpx_read_mps, lpx_read_freemps, lpx_write_mps,
481        lpx_write_freemps, lpx_read_cpxlp, and lpx_write_cpxlp.
483        Two bugs were fixed. Thanks to
484        Anne-Laurence Putz <> and
485        Xypron <> for bug report.
487GLPK 4.28 (release date: Mar 25, 2008)
489        The iODBC and MySQL table drivers, which allows transmitting
490        data between MathProg model objects and relational databases,
491        were re-implemented to replace a static linking by a dynamic
492        linking to corresponding shared libraries.
493        Many thanks to Heinrich Schuchardt <>
494        for the contribution, Rafael Laboissiere <>
495        for useful advices concerning the shared library support under
496        GNU/Linux, and Vijay Patil <> for testing
497        this feature under Windows XP.
499        A new optional feature was added to the package. This feature
500        is based on the zlib data compression library and allows GLPK
501        API routines and the stand-alone solver to read and write
502        compressed data files performing compression/decompression "on
503        the fly" (compressed data files are recognized by suffix `.gz'
504        in the file name). It may be useful in case of large MPS files
505        to save the disk space (up to ten times).
507        The `configure' script was re-implemented. Now it supports the
508        following specific options:
510        --with-gmp           Enable using the GNU MP bignum library
511        --without-gmp        Disable using the GNU MP bignum library
512        --with-zlib          Enable using the zlib data compression
513                             library
514        --without-zlib       Disable using the zlib data compression
515                             library
516        --enable-dl          Enable shared library support (auto check)
517        --enable-dl=ltdl     Enable shared library support (GNU)
518        --enable-dl=dlfcn    Enable shared library support (POSIX)
519        --disable-dl         Disable shared library support
520        --enable-odbc        Enable using ODBC table driver
521        --disable-odbc       Disable using ODBC table driver
522        --enable-mysql       Enable using MySQL table driver
523        --disable-mysql      Disable using MySQL table driver
525        For more details please see file INSTALL.
527GLPK 4.27 (release date: Mar 02, 2008)
529        Three new table drivers were added to the MathProg translator:
531        xBASE built-in table driver, which allows reading and writing
532        data in .dbf format (only C and N fields are supported);
534        MySQL table driver, which provides connection to a MySQL
535        database;
537        iODBC table driver, which provides connection to a database
538        through ODBC.
540        The MySQL and iODBC table drivers were contributed to GLPK by
541        Heinrich Schuchardt <>.
543        The table driver is a program module which allows transmitting
544        data between MathProg model objects and external data tables.
546        For detailed description of the table statement and table
547        drivers see the document "Using Data Tables in the GNU MathProg
548        Modeling Language" (file doc/tables.txt) included in the
549        distribution. Some examples which demonstrate using MySQL and
550        iODBC table drivers can be found in subdirectory examples/sql.
552GLPK 4.26 (release date: Feb 17, 2008)
554        The table statement was implemented in the GNU MathProg
555        modeling language. This new feature allows reading data from
556        external tables into model objects such as sets and parameters
557        as well as writing results of computations to external tables.
559        A table is a (unordered) set of records, where each record
560        consists of the same number of fields, and each field is
561        provided with a unique symbolic name called the field name.
563        Currently the GLPK package has the only built-in table driver,
564        which supports tables in the CSV (comma-separated values) file
565        format. This format is very simple and supported by almost all
566        spreadsheets and database management systems.
568        Detailed description of the table statement and CSV format can
569        be found in file doc/tables.txt, included in the distribution.
571GLPK 4.25 (release date: Dec 19, 2007)
573        A tentative implementation of Gomory's mixed integer cuts was
574        included in the branch-and-cut solver. To enable generating
575        Gomory's cuts the control parameter gmi_cuts passed to the
576        routine glp_intopt should be set to GLP_ON. This feature is
577        also available in the solver glpsol through command-line option
578        '--gomory'. For more details please see the reference manual
579        included in the distribution.
581GLPK 4.24 (release date: Nov 21, 2007)
583        A tentative implementation of MIR (mixed integer rounding) cuts
584        was included in the MIP solver. To enable generating MIR cuts
585        the control parameter mir_cuts passed to the routine glp_intopt
586        should be set to GLP_ON. This feature is also available in the
587        stand-alone solver glpsol via command-line option '--mir'. For
588        more details please see the reference manual included in the
589        distribution.
591        The implementation is mainly based on the following two papers:
593        1. H. Marchand and L. A. Wolsey. Aggregation and mixed integer
594           rounding to solve MIPs. CORE discussion paper 9839, CORE,
595           Universite catholique de Louvain, June 1998.
597        2. G. Andreello, A. Caprara, and M. Fischetti. Embedding cuts
598           in a Branch&Cut framework. Preliminary draft, October 2003.
600        MIR cuts can be generated on any level of the search tree that
601        makes the GLPK MIP solver to be a real branch-and-cut solver.
603        A bug was fixed in the routine lpx_write_cpxlp. If a variable
604        x has upper bound and no lower bound, it should appear in the
605        bounds section as "-inf <= x <= u", not as "x <= u". Thanks to
606        Enric Rodriguez <> for the bug report.
608GLPK 4.23 (release date: Oct 28, 2007)
610        The following new API routines were added:
612        glp_read_sol    read basic solution from text file
613        glp_write_sol   write basic solution to text file
614        glp_read_ipt    read interior-point solution from text file
615        glp_write_ipt   write interior-point solution to text file
616        glp_read_mip    read MIP solution from text file
617        glp_write_mip   write MIP solution to text file
619        For description of these routines and corresponding file
620        formats see Chapter "API Routines", Section "Utility routines"
621        in the reference manual included in the distribution.
623        Advanced API routine glp_free_env was added. It may be used by
624        the application program to free all resources allocated by GLPK
625        routines.
627        The following three new command-line options were added to the
628        solver glpsol:
630        --mipgap tol    set relative MIP gap tolerance
631        -r filename     read solution from filename
632        -w filename     write solution to filename
634GLPK 4.22 (release date: Sep 19, 2007)
636        This is a maintainer release.
638        A bug was fixed in the MIP preprocessor (ios_preprocess_node).
639        Thanks to Roberto Bagnara <> (Department of
640        Mathematics, University of Parma, Italy) for the bug report.
642        A bug was fixed in the MIP preprocessor (col_implied_bounds),
643        due to which constraint coefficients with small magnitude could
644        lead to wrong implied bounds of structural variables.
646        A similar bug was fixed in the routine reduce_bounds.
648        A bug was fixed in the routines glp_set_mat_row and
649        glp_set_mat_col. (The bug appeared due to incorrect removing
650        zero elements from the row/column lists.)
652        A bug was fixed in the API routines lpx_read_mps and
653        lpx_read_freemps, due to which bounds of type LI specified in
654        BOUNDS section were incorrectly processed.
656        A call to standard function vsprintf was replaced by a call to
657        vsnprintf for security reasons. Many thanks to Peter T. Breuer
658        <> and Rafael Laboissiere <>.
660GLPK 4.21 (release date: Aug 28, 2007)
662        Additional reasons for calling the callback routine used in the
663        MIP solver (glp_intopt) were introduced. Currently the following
664        reasons are supported:
666        * request for subproblem selection
667        * request for preprocessing
668        * request for row generation
669        * request for heuristic solution
670        * request for cut generation
671        * request for branching
672        * better integer solution found
674        A basic preprocessing component used to improve subproblem
675        formulations by tightening bounds of variables was included in
676        the MIP solver. Depending on the control parameter pp_tech
677        passed to the routine glp_intopt the preprocessing can be
678        performed either on the root level or on all levels (default)
679        or can be disabled.
681        Backtracking heuristic used by default in the MIP solver was
682        changed to the "best local bound".
684        For more details see Chapter "Advanced API routines", Section
685        "Branch-and-bound interface routines" in a new edition of the
686        reference manual included in the distribution.
688GLPK 4.20 (release date: Jul 26, 2007)
690        API routine lpx_integer was replaced by API routine glp_intopt,
691        which provides equivalent functionality and additionally allows
692        the application to control the solution process by means of the
693        user-written callback routine, which is called by the solver at
694        various points of the branch-and-bound algorithm. Besides, the
695        new MIP solver allows generating "lazy" constraints and cutting
696        planes on all levels of the branch-and-bound tree, not only on
697        the root level. The routine lpx_integer is also still available
698        for the backward compatibility.
700        The following new advanced API routines, which may be called
701        from the B&B callback routine, were included in the package:
703        glp_ios_reason     determine reason for calling callback
704                           routine
705        glp_ios_get_prob   access the problem object
706        glp_ios_tree_size  determine size of the branch-and-bound tree
707        glp_ios_curr_node  determine current active subproblem
708        glp_ios_next_node  determine next active subproblem
709        glp_ios_prev_node  determine previous active subproblem
710        glp_ios_up_node    determine parent subproblem
711        glp_ios_node_level determine subproblem level
712        glp_ios_node_bound determine subproblem local bound
713        glp_ios_mip_gap    compute relative MIP gap
714        glp_ios_heur_sol   provide solution found by heuristic
715        glp_ios_terminate  terminate the solution process
717        For description of these routines see Chapter "Advanced API
718        routines", Section "Branch-and-bound interface routines" in a
719        new edition of the reference manual, which was included in the
720        distribution.
722        Old version of the integer optimization suite (IOS) as well as
723        TSP solver tspsol based on it are no longer supported and were
724        removed from the package.
726        A minor error in the MIP presolver was fixed; thanks to Graham
727        Rockwell <> for the bug report.
729GLPK 4.19 (release date: Jul 05, 2007)
731        The principal change is upgrading to GPLv3.
733        A serious bug in the routine glp_del_cols was fixed; thanks to
734        Cedric[FR] <> for the bug report. The bug
735        appeared because on deleting non-basic columns the basis header
736        remained valid, however, contained invalid (old) column ordinal
737        numbers.
739        A new advanced API routine glp_mem_limit was added.
741        The case GLP_EBOUND was added to the routine lpx_simplex.
742        Thanks to Cameron Kellough <> for the
743        bug report.
745        An API routine lpx_write_pb to write the problem instance in
746        OPB (pseudo boolean) format format was added. Thanks to Oscar
747        Gustafsson <> for the contribution.
749        Two new options --wpb and --wnpb were added to glpsol to write
750        the problem instance in OPB format.
752GLPK 4.18 (release date: Jun 25, 2007)
754        The following new API routines were added:
756        glp_set_rii        set (change) row scale factor
757        glp_set_sjj        set (change) column scale factor
758        glp_get_rii        retrieve row scale factor
759        glp_get_sjj        retrieve column scale factor
760        glp_simplex        solve LP problem with the simplex method
761                           (this routine replaces lpx_simplex, which is
762                           also available for backward compatibility)
763        glp_init_smcp      initialize simplex method control params
764        glp_bf_exists      check if the basis factorization exists
765        glp_factorize      compute the basis factorization
766        glp_bf_updated     check if the basis factorization has been
767                           updated
768        glp_get_bfcp       retrieve basis factorization control params
769        glp_set_bfcp       change basis factorization control params
770        glp_get_bhead      retrieve the basis header information
771        glp_get_row_bind   retrieve row index in the basis header
772        glp_get_col_bind   retrieve column index in the basis header
773        glp_ftran          perform forward transformation
774        glp_btran          perform backward transformation
776        For description of all these routines see a new edition of the
777        reference manual included in the distribution.
779        Type names ulong_t and uldiv_t were changed to glp_ulong and
780        glp_uldiv to avoid conflicts with standard type names on some
781        platforms. Thanks to Boris Wirtz <>
782        for the bug report.
784        Some new examples in the MathProg language were added. Thanks
785        to Sebastian Nowozin <>.
787GLPK 4.17 (release date: May 26, 2007)
789        API routines glp_set_mat_row, glp_set_mat_col, and glp_load_mat
790        were modified to allow zero constraint coefficients (which are
791        not stored in the constraint matrix). Note that constraint
792        coefficients with duplicate row/column indices are not allowed.
794        Another form of LP basis factorization was implemented in the
795        package. It is based on LU-factorization of an initial basis
796        and Schur complement to reflect changes in the basis. Currently
797        the implementation is incomplete and provides only updating the
798        factorization on replacing a column of the basis matrix. On API
799        level the user can set the control parameter LPX_K_BFTYPE to
800        choose between the folloiwng forms of LP basis factorization to
801        be used in the simplex method routines:
802        1) LU + Forrest-Tomlin update;
803        2) LU + Schur complement + Bartels-Golub update;
804        3) LU + Schur complement + Givens rotation update.
805        The GLPK implementation is similar to LUSOL/LUMOD developed by
806        Michael A. Saunders.
808        The user can choose the form of LP basis factorzation used by
809        the simplex method routines by specifying the folloiwng options
810        of glpsol: --luf, --cbg, --cgr.
812GLPK 4.16 (release date: May 05, 2007)
814        A number of basic GLPK API routines, which now are in the
815        stable stable, were renamed to be prefixed with 'glp_'. Note
816        that all these routines are available via their old names
817        prefixed with 'lpx_' that keeps the downward compatibility with
818        older versions of the package.
820        Three new GLPK API routines were added to the package:
821        glp_version, glp_term_hook, and glp_mem_usage; for more details
822        see a new edition of the GLPK reference manual included in the
823        distribution. The routine glp_version reports the actual version
824        of the GLPK library and also can be used (along with the header
825        glpk.h) in Autotools specification files to check if the GLPK
826        library has been installed.
828        The header glpk.h was changed to conform to C++ environment.
830GLPK 4.15 (release date: Feb 18, 2007)
832        Autotools specification files (, were
833        changed to use GNU Libtool. This allows building the static as
834        well as shared GLPK library.
836GLPK 4.14 (release date: Feb 05, 2007)
838        Now GLPK conforms to ILP32, LLP64, and LP64 programming models
839        (the latter seems to be the ultimate choice regarding 64-bit
840        architectures). Note that GLPK itself is a 32-bit application,
841        and the conformity only means that the package works correctly
842        on all these arenae. Nevertheless, on 64-bit platforms it is
843        possible to use more than 4GB of memory, if necessary.
845GLPK 4.13 (release date: Nov 13, 2006)
847        A tentative implementation of the "exact" simplex method based
848        on bignum (rational) arithmetic was included in the package.
850        On API level this new feature is available through the routine
851        lpx_exact, which is similar to the routine lpx_simplex.
853        In the solver glpsol this feature is available through two new
854        command-line options: --exact and --xcheck. If the '--exact'
855        option is specified, glpsol solves LP instance using the exact
856        simplex method; in case of MIP it is used to obtain optimal
857        solution of LP relaxation. If the --xcheck option is specified,
858        LP instance (or LP relaxation) is solved using the standard
859        (floating-point) simplex method, however, then glpsol calls the
860        exact simplex routine to make sure that the final LP basis is
861        exactly optimal, and if it is not, to perform some additional
862        simplex iterations in exact arithmetic.
864GLPK 4.12 (release date: Nov 08, 2006)
866        A tentative implementation of some simplex method routines
867        based on exact (bignum) arithmetic was included in the package.
868        Currently these routines provide computing LU-factorization of
869        the basis matrix and computing components of basic solution.
871        These routines were used to implement a routine, which checks
872        primal and dual feasibility of basic solution exactly, i.e. in
873        rational numbers, without round-off errors. In glpsol this
874        feature is available through the command-line option --xcheck.
876        GLPK has its own low-level routines implementing operations on
877        integer and rational numbers that makes it independent on other
878        software packages. However, to attain a much better performance
879        it is highly recommended to install (before configuring GLPK)
880        the GNU Multiple Precision Arithmetic Library (GMP). Using GMP
881        makes computations 100-200 times faster.
883GLPK 4.11 (release date: Jul 25, 2006)
885        Three new built-in functions in the modeling language were
886        implemented: card (cardinality of set), length (length of
887        character string), and substr (substring of character string).
888        Another improvement concerns the printf statement which now
889        allows redirecting its output to a specified file. These new
890        features are illustrated in example models crypto.mod and
891        graph.mod included in the distribution. For more details see
892        the document "Modeling Language GNU MathProg".
894        Four batch files (along with corresponding makefiles) were
895        included in the distribution to simplify building GLPK under
896        MS Windows; see them in subdirectory 'w32'.
898GLPK 4.10 (release date: May 11, 2006)
900        Cutting planes of two new classes were implemented: mixed cover
901        cuts and clique cuts. On API level this feature can be enabled
902        by setting control parameter LPX_K_USECUTS passed to the routine
903        lpx_intopt. In glpsol this feature is available through the
904        command-line options --cover and --clique. For more details see
905        the reference manual.
907        Now the routines lpx_read_mps and lpx_read_freemps support LI
908        bound type. It is similar to LO, however, indicates the column
909        as of integer kind.
911GLPK 4.9 (release date: Jan 17, 2006)
913        An advanced MIP solver was implemented. It includes:
915        - basic presolving technique (removing free, singleton and
916          redundant rows, improving bounds of columns, removing fixed
917          columns, reducing constraint coefficents);
919        - generating cutting planes to improve LP relaxation (currently
920          only Gomory's mixed integer cuts are implemented);
922        - using the branch-and-bound method to solve resultant MIP;
924        - recovering solution of the original MIP.
926        The solver is available on API level via the routine lpx_intopt
927        (see the reference manual). It is similar to the routine
928        lpx_integer, however, does not require initial solution of LP
929        relaxation.
931        The solver is also available in the command-line utility glpsol
932        via two options: --intopt (only presolving) and --cuts (assumes
933        --intopt plus generating cuts).
935        Note that efficiency of the MIP solver strongly depends on the
936        internal structure of the problem to be solved. For some hard
937        instances it is very efficient, but for other instances it may
938        be significantly worse than the standard branch-and-bound.
940        For some comparative benchmarks see doc/bench1.txt.
942        Well, what else...
944        Three built-in functions were added to MathProg: sin, cos, and
945        atan (the latter allows one or two arguments).
947        Some bugs were fixed.
949        Several new examples in MathProg were included: color.mod
950        (graph coloring problem), tsp.mod (traveling salesman problem),
951        and pbn.mod (paint-by-numbers puzzle).
953GLPK 4.8 (release date: Jan 12, 2005)
955        Core simplex method and interior-point method routines were
956        re-implemented and now they use a new, "storage-by-rows" sparse
957        matrix format (unlike previous versions where linked lists were
958        used to represent sparse matrices). For details see ChangeLog.
960        Also a minor bug was fixed in API routine lpx_read_cpxlp.
962GLPK 4.7 (release date: Aug 23, 2004)
964        Now GLPK supports free MPS format. Two new API routines
965        lpx_read_freemps (to read problem data in free MPS format) and
966        lpx_write_freemps (to write problem data in free MPS format)
967        were added. This feature is also available in the solver glpsol
968        via new command-line options --freemps and --wfreemps. For more
969        details see the GLPK reference manual.
971        API routines lpx_read_cpxlp and lpx_write_cpxlp for reading and
972        writing problem data in CPLEX LP format were re-implemented to
973        allow long symbolic names (up to 255 characters).
975        The following three modules were temporarily removed from the
976        GLPK distribution due to licensing problems: DELI (an interface
977        module to Delphi), GLPKMEX (an interface module to Matlab), and
978        JNI (an interface module to Java).
980GLPK 4.6 (release date: Aug 04, 2004)
982        Three new statements were implemented in the GNU MathProg
983        language: solve, printf, and for. Their detailed description can
984        be found in the GLPK documentation included in the distribution.
985        (See also a sample model, examples/queens.mod, which illustrates
986        using these new statements.)
988        Two new API routines were added to the package: lpx_read_prob
989        and lpx_write_prob. They allow reading/writing problem data in
990        GNU LP low-level text format.
992        Three new command-line options were implemented in the LP/MIP
993        solver glpsol: --glp (to read problem data in GNU LP format),
994        --wglp (to write problem data in GNU LP format), and --name (to
995        change problem name). Now glpsol also supports processing models
996        where the new statements (see above) are used.
998        A new version of GLPKMEX, a Matlab MEX interface to GLPK, was
999        included. For more details see contrib/glpkmex/ChangeLog.
1001GLPK 4.5 (release date: Jul 19, 2004)
1003        The branch-and-bound solver was completely re-implemented.
1005        Some modifications were made in memory allocation routines that
1006        allows using the package on 64-bit platforms.
1008        For more details see ChangeLog.
1010GLPK 4.4 (release date: Jan 17, 2004)
1012        All API routines were re-implemented using new data structures.
1013        The new implementation provides the same specifications and
1014        functionality of API routines as the old one, however, it has
1015        some important advantages, in particular:
1016        * linked lists are used everywhere that allows creating and
1017          modifying the problem object as efficiently as possible
1018        * all data stored in the problem object are non-scaled (even if
1019          the internal scaling is used) that prevents distortion of the
1020          original problem data
1021        * solution components obtained by the solver remain available
1022          even if the problem object has been modified
1023        * no solver-specific data are used in the new data structures
1024          that allows attaching any external lp/mip solver using GLPK
1025          API as an uniform interface
1026        Note that some API routines became obsolete being replaced by
1027        new, more convenient routines. These obsolete routines are kept
1028        for backward compatibility, however, they will be removed in
1029        the future. For more details please see ChangeLog and the GLPK
1030        Reference Manual.
1032        New edition of the GLPK Reference Manual was included in the
1033        distribution.
1035        GLPKMEX, a Matlab MEX interface to GLPK package, contributed by
1036        Nicolo Giorgetti <> was included in the
1037        distribution.
1039        GLPK FAQ contributed by Harley Mackenzie <> was
1040        included in the distribution.
1042GLPK 4.3 (release date: Dec 12, 2003)
1044        The bug, due to which the standard math library is not linked
1045        on building the package on some platforms, was fixed.
1047        The following new built-in functions were added to the MathProg
1048        language: round, trunc, Irand224, Uniform01, Uniform, Normal01,
1049        Normal. For details see the language description.
1051        The MathProg syntax was changed to allow writing 'subj to' that
1052        means 'subject to'.
1054        The new api routine lpx_get_ray_info was added. It is intended
1055        to determine which (non-basic) variable causes unboundness. For
1056        details see the reference manual.
1058        The module glpmps.c was changed to avoid compilation errors on
1059        building the package on Mac OS X.
1061        Several typos was fixed and some new material was added to the
1062        GLPK documentation.
1064GLPK 4.2 (release date: Nov 14, 2003)
1066        A preliminary implementation of the Integer Optimization Suite
1067        (IOS) was included in the package. The Branch-and-Cut Framework
1068        being completely superseded by IOS was removed from the package.
1070        New API routine lpx_print_sens_bnds intended for bounds
1071        sensitivity analysis was contributed to GLPK by Brady Hunsaker
1072        <>. This function is also available in
1073        the solver glpsol (via command-line option --bounds).
1075        An improved version of GLPK JNI (Java Native Interface) was
1076        contributed by Chris Rosebrugh <>.
1078        GLPK DELI (Delphi Interface) was contributed by Ivo van Baren
1079        <>.
1081        Several makefiles were added to allow compiling GLPK on some
1082        non-GNU 32-bit platforms:
1083        * Windows single-threaded static library, Visual C++ 6.0
1084        * Windows multi-threaded dynamic library, Visual C++ 6.0
1085        * Windows single-threaded static library, Borland C++ 5.2
1086        * DOS single-threaded static library, Digital Mars C++ 7.50
1088        And, of course, some bugs were fixed.
1090        For more details see ChangeLog.
1092GLPK 4.1 (release date: Aug 23, 2003)
1094        Some improvements were made in the lp/mip solver routines and
1095        several bugs were fixed in the model translator.
1097        For more details see ChangeLog.
1099GLPK 4.0 (release date: May 06, 2003)
1101        Now GLPK supports the GNU MathProg modeling language, which is
1102        a subset of the AMPL modeling language.
1104        The document "GLPK: Modeling Language GNU MathProg" included in
1105        the distribution is a complete description of GNU MathProg. (See
1106        the files lang.latex, lang.dvi, and in the subdirectory
1107        'doc'. See also some examples in the subdirectory 'sample'.)
1109        New version of the solver glpsol, which supports models written
1110        in GNU MathProg, was implemented. (Brief instructions how to use
1111        glpsol can be found in the GNU MathProg documentation.)
1113        The GLPK/L modeling language is no more supported. The reason is
1114        that GNU MathProg being much more powerful completely supersedes
1115        all features of GLPK/L.
1117GLPK 3.3 (release date: Mar 25, 2003)
1119        LP PRESOLVER
1120        ------------
1122        Now the routine lpx_simplex (which is a driver to the simplex
1123        method for solving LP) is provided with the built-in LP
1124        presolver, which is a program that transforms the original LP
1125        problem to an equivalent LP problem, which may be easier for
1126        solving with the simplex method than the original one. Once the
1127        transformed LP has been solver, the presolver transforms its
1128        basic solution back to a corresponding basic solution of the
1129        original problem. For details about this feature please see the
1130        GLPK reference manual.
1132        Currently the LP presolver implements the following features:
1133        * removing empty rows;
1134        * removing empty columns;
1135        * removing free rows;
1136        * removing fixed columns;
1137        * removing row singletons, which have the form of equations;
1138        * removing row singletons, which have the form of inequalities;
1139        * removing column singletons, which are implied slack variables;
1140        * fixing and removing column singletons, which are implied free
1141          variables;
1142        * removing forcing rows that involves fixing and removing the
1143          corresponding columns;
1144        * checking for primal and dual infeasibilities.
1146        The LP presolver is also used by default in the stand-alone
1147        program glpsol. In order *not* to use it, the option --nopresol
1148        should be specified in the command-line.
1150        CHANGES IN GLPK/L
1151        -----------------
1153        The syntax and semantics of the GLPK/L modeling language was
1154        changed to allow declaration of "interval" sets. This means that
1155        now the user can declare a set, for example, as:
1157           set task = [8:11];
1159        that is exactly equivalent to the following declaration:
1161           set task = (task_8, task_9, task_10, task_11);
1163        For details see the language description.
1165        JAVA INTERFACE
1166        --------------
1168        Now GLPK includes the package GLPK JNI (Java Native Interface)
1169        that implements Java binding for GLPK. It allows Java programs
1170        to utilize GLPK in solving LP and MIP problems. For details see
1171        a brief user's guide in the subdirectory contrib/java-binding.
1172        This package was developed and programmed by Yuri Victorovich
1173        <>, who contributed it to GLPK.
1175GLPK 3.2.4 (release date: Feb 18, 2003)
1177        This is a bug-fix release. For details see ChangeLog.
1179GLPK 3.2.3 (release date: Nov 11, 2002)
1181        A new implementation of the api routine lpx_integer which now
1182        is based on the b&b driver (which is based on the implicit
1183        enumeration suite) was included in the package. This new
1184        implementation has exactly the same functionality as the old
1185        version, so all changes are transparent to the api user.
1187        Four new api routines were included in the package:
1188        lpx_check_kkt checks Karush-Kuhn-Tucker optmality conditions;
1189        lpx_read_bas reads predifined basis in MPS format;
1190        lpx_write_bas writes current basis in MPS format;
1191        lpx_write_lpt writes problem data in CPLEX LP format.
1193        Also other minor improvements were made (for details see the
1194        file 'ChangeLog').
1196GLPK 3.2.2 (release date: Oct 14, 2002)
1198        The api routine lpx_read_lpt was included in the package. It
1199        is similar to the routine lpx_read_mps and intended to read
1200        LP/MIP data prepared in CPLEX LP format. Description of this
1201        format is given in the GLPK reference manual, a new edition of
1202        which was also included in the distribution (see the files
1203        'refman.latex', 'refman.dvi', '' in the subdirectory
1204        'doc'). In order to use data files in CPLEX LP format with the
1205        solver glpsol the option '--lpt' should be specified in the
1206        command line.
1208        Several bugs were fixed and some minor improvements were made
1209        (for details see the file 'ChangeLog').
1211GLPK 3.2.1 (release date: Aug 12, 2002)
1213        Now GLPK includes a preliminary implementation of the
1214        branch-and-cut framework, which is a set of data structures and
1215        routines intended for developing branch-and-cut methods for
1216        solving mixed-integer and combinatorial optimization problems.
1218        Detailed decsription of the branch-and-cut framework is given in
1219        the document "GLPK: A Preliminary Implementation of the
1220        Branch-And-Cut Framework" included in the distribution (see the
1221        file 'brcut.txt' in the subdirectory 'doc').
1223        In order to illustrate how the GLPK branch-and-cut framework
1224        can be used for solving a particular optimization problem there
1225        is an example included in the package. This is a stand-alone
1226        program, TSPSOL, which is intended for solving to optimality the
1227        symmetric Traveling Salesman Problem (TSP), a classical problem
1228        of the combinatorial optimization (see the file 'tspsol.c' in
1229        the subdirectory 'sample').
1231GLPK 3.2 (release date: Jul 15, 2002)
1233        New edition of the document "GLPK: Reference Manual" was
1234        included (see the files 'refman.latex', 'refman.dvi', and
1235        '' in the subdirectory 'doc').
1237        New edition of the document "GLPK: Modeling Language GLPK/L" was
1238        included (see the files 'lang.latex', 'lang.dvi', and ''
1239        in the subdirectory 'doc').
1241        The following new API routines were added to the package:
1243        lpx_transform_row (transform explicitly specified row);
1244        lpx_transform_col (transform explicitly specified column);
1245        lpx_prim_ratio_test (perform primal ratio test);
1246        lpx_dual_ratio_test (perform dual ratio test);
1247        lpx_interior (solve LP problem using interior point method);
1248        lpx_get_ips_stat (query status of interior point solution);
1249        lpx_get_ips_row (obtain row interior point solution);
1250        lpx_get_ips_col (obtain column interior point solution);
1251        lpx_get_ips_obj (obtain interior point value of obj.func.);
1252        lpx_read_lpm (read LP/MIP model written in GLPK/L);
1253        lpx_write_mps (write problem data using MPS format);
1254        lpx_print_ips (print interior point solution).
1256        Detailed description of all these new API routines are given in
1257        the new edition of the reference manual.
1259        New version of the stand-alone solver glpsol (which is based on
1260        the new API) was implemented.
1262        So long as the new API (introduced in glpk 3.0) now provides
1263        all the functions, which were provided by the old API, the old
1264        API routines were removed from the package at all.
1266GLPK 3.1 (release date: May 27, 2002)
1268        A preliminary implementation of new API routines was completed
1269        and included in the package.
1271        These new API routines provide much more flexible interaction
1272        between the application program, LP/MIP problem instances, and
1273        solver routines. Based on completely changed data structures
1274        they are, however, similar to the API routines and provide the
1275        same functionality. Please note that three routines, namely,
1276        solving LPs using interior point method, reading model written
1277        in the GLPK/L modeling language, and writing problem data in
1278        the MPS format, are not implemented in the new API, however,
1279        these routines are planned to be implemented in the next version
1280        of the package.
1282        A description of the new API routines is given in the document
1283        "GLPK Reference Manual", a draft edition of which is included
1284        in the package (see the files 'refman.latex', 'refman.dvi', and
1285        '' in the subdirectory 'doc').
1287        Although the old API routines are kept in the package, they are
1288        no longer supported and will be removed in the future.
1290GLPK 3.0.8 (release date: May 13, 2002)
1292        A preliminary implementation of new API routines was included
1293        in the package. These new API routines are intended to provide
1294        much more flexible interaction between the application program,
1295        LP/MIP problem and solver routines. See the document "New GLPK
1296        API Routines" (the file 'newapi.txt' in the subdirectory 'doc')
1297        also included in the package.
1299        The api routines glp_simplex2, glp_call_ipm1, glp_call_bbm1 were
1300        renamed, respectively, to glp_simplex, glp_interior, glp_integer
1301        in order to reflect changes in implementation. The api routines
1302        glp_call_rsm1, glp_simplex1, glp_pivot_in, glp_pivout_out were
1303        removed from the package since they are completely supreseded by
1304        the new API routines (however, these routines still can be found
1305        in the subdirectory 'oldsrc'). Please consult a new edition of
1306        the document "GLPK User's Guide" about all these changes in the
1307        existing api routines.
1309        The document "GLPK Library Reference" was removed from the
1310        package (into the subdirectory 'oldsrc') since it describes the
1311        obsolete library routines, most of which are no longer used.
1313GLPK 3.0.7 (release date: Apr 22, 2002)
1315        A new, more efficient implementation of the primal/dual simplex
1316        method was included in the package. Due to some improvements the
1317        simplex-based solver allows solving many LP problems faster and
1318        provides more reliable results. Note that the new implementation
1319        is currently incomplete and available only via the api routine
1320        glp_simplex2.
1322        All the changes are transparent on API level.
1324GLPK 3.0.6 (release date: Mar 28, 2002)
1326        New version of LU-factorization and basis maintenance routines
1327        (based on Forrest-Tomlin updating technique) was implemented.
1328        Since these new routines functionally supersede some routines
1329        (which implement other forms of the basis matrix) and make them
1330        obsolete, the latter were removed from the package (they still
1331        can be found in the subdirectory 'oldsrc').
1333        All the changes are transparent on API level.
1335GLPK 3.0.5 (release date: Jan 29, 2002)
1337        New edition of the document "GLPK User's Guide" was included in
1338        the distribution. Now it describes all additional API routines,
1339        which were recently added to the package.
1341        Structure of the package was re-organized in order to make its
1342        maintenance easier (all small files in the subdurectory 'source'
1343        were merged in bigger units). These changes are transparent for
1344        the user.
1346GLPK 3.0.4 (release date: Dec 10, 2001)
1348        A new, more efficient implementation of the two-phase primal
1349        simplex method was included in the package. Due to some new
1350        features (an advanced initial basis, projected steepest edge,
1351        recursive updating values and reduced costs) the new LP solver
1352        is faster and numerically more stable than the old one.
1354        The new LP solver is available as API routine glp_simplex2 and
1355        has the same purpose as API routine glp_call_rsm1. For detailed
1356        specification see the file 'newapi.txt' in the directory 'doc'.
1358        Now the new LP solver is also used by default to solve an
1359        initial LP problem in the branch-and-bound routine glp_call_bbm1
1360        instead the routine rsm1_driver. Note that the branch-and-bound
1361        procedure itself is still based on rsm1_driver.
1363        The new LP solver is also used as default solver in GLPSOL for
1364        solving LP and MIP problems. In order to choose the old solver
1365        the option '--old-sim' can be specified in the command line.
1367GLPK 3.0.3 (release date: Oct 03, 2001)
1369        Some minor changes were made in the simplex method routines in
1370        order to improve numerical stability of the method.
1372GLPK 3.0.2 (release date: Sep 24, 2001)
1374        A new implementation of the basis maintaining routines was
1375        included in the package. These routines, which are based on so
1376        called FHV-factorization (a variety of LU-factorization) of the
1377        basis matrix and Gustavson's data structures, allows performing
1378        the main operations faster at the expense of some worsening
1379        numerical accuracy.
1381        AFI (Advanced Form of the Inverse), which is the form of the
1382        basis matrix based on FHV-factorization, is available via the
1383        parameter form = 3 (on API level) or via the option --afi (in
1384        GLPSOL solver).
1386GLPK 3.0.1 (release date: Aug 01, 2001)
1388        Old GLPK API routines have been removed from the package.
1390        New GLPK API routines were added:
1392        - scaling routines;
1394        - a routine for writing problem data in MPS format;
1396        - a comprehensive driver to the simplex method;
1398        - basis maintaining routines.
1400        A description of the new API routines is given in the document
1401        "Additional GLPK API Routines". This document is included into
1402        the distribution in plain text format (see the file 'newapi.txt'
1403        in the subdirectory 'doc').
1405        Now the distribution includes a non-trivial example of using
1406        GLPK as a base LP solver for Concorde, a well known program that
1407        solves Traveling Salesman Problem (TSP). For further details see
1408        comments in the file 'sample/lpglpk30.c'.
1410GLPK 3.0 (release date: Jul 19, 2001)
1412        Now GLPK is provided with new API, which being more flexible
1413        can be used in more complex algorithmic schemes.
1415        New edition of the document "GLPK User's Guide" is included in
1416        the distribution. Now it completely corresponds to the new GLPK
1417        API routines.
1419        Old API routines are not removed yet from the package, however
1420        they became obsolete and therefore should not be used. Since now
1421        the header glpk.h corresponds to new API, in order to compile
1422        existing programs that use old GLPK API routines the statement
1424        #define GLP_OLD_API
1426        should be inserted before the statement
1428        #include "glpk.h"
1430GLPK 2.4.1 (release date: Jun 14, 2001)
1432        The document "Modeling language GLPK/L" is included into the
1433        distribution in texinfo format.
1435        New edition of the document "GLPK User's Guide" is included in
1436        the distribution. Now it describes all additional API routines
1437        which were recently added to the package.
1439GLPK 2.4 (release date: May 10, 2001)
1441        Now GLPK includes an implementation of a preliminary version
1442        of the GLPK/L modeling language. This language is intended for
1443        writing mathematcal programming models. The name GLPK/L is
1444        derived from GNU Linear Programming Kit Language.
1446        A brief description of the GLPK/L language is given in the
1447        document "GLPK/L Modeling Language: A Brief Description". This
1448        document is included into the distribution in plain text format
1449        (see the file 'language.txt' in the subdirectory 'doc').
1451        The language processor (which is a program that analyzes model
1452        description written in GLPK/L and translates it to internal data
1453        structures) is available as the GLPK API routine.
1455        The stand-alone solver GLPSOL now is able: a) to process model
1456        descriptions written in the GLPK/L language; b) to solve pure LP
1457        problems using the interior point method (therefore the program
1458        GLPIPM was removed from the package).
1460GLPK 2.3 (release date: Apr 09, 2001)
1462        New edition of the document "GLPK User's Guide" is included in
1463        the distribution. Now it describes all additional API routines
1464        which were recently added to the package.
1466        The MIP solver was fully re-programmed in order to improve its
1467        robustness and performance. In particular, a basis recovering
1468        procedure was implemented (this procedure allows switching to
1469        the primal simplex method in case when the dual simplex method
1470        fails).
1472GLPK 2.2 (release date: Mar 15, 2001)
1474        Now GLPK includes a tentative implementation of the
1475        branch-and-bound procedure based on the dual simplex method for
1476        mixed integer linear programming (MIP).
1478        Complete description of this new feature of the package is given
1479        in the preliminary document "Mixed Integer Linear Programming
1480        Using GLPK Version 2.2 (Supplement to GLPK User's Guide)". This
1481        document is included into the distribution in plain text format
1482        (see the file 'mip.txt' in the subdirectory 'doc').
1484        The MIP solver (glp_integer) can be used as GLPK API routine in
1485        the same way as the pure LP solver (glp_simplex).
1487        The stand-alone program 'glpsol' is now able to solve LP as well
1488        as MIP problems.
1490        Note that the current version of GLPK MIP solver is based on
1491        easiest heuristics for branching and backtrackng. Therefore the
1492        solver is fit mainly for MIP problems which are not very hard
1493        and have few integer variables.
1495GLPK 2.1 (release date: Feb 19, 2001)
1497        The document "GLPK Implementation of the Revised Simplex Method"
1498        is included into the distribution. This document describes most
1499        of routines related to the revised simplex method.
1501GLPK 2.0 (release date: Jan 25, 2001)
1503        Now GLPK includes a tentative implementation of the primal-dual
1504        interior point method for large-scale linear programming.
1506        The interior point solver can be used as GLPK API routine in the
1507        same manner as the simplex method solver (glp_simplex):
1509        ret = glp_interior();
1511        Note that currently the interior point solver implemented in
1512        GLPK doesn't include many important features, in particular:
1514        * it can't process dense columns; therefore if the problem has
1515          dense columns, the solving will be extremely inefficient;
1517        * it has no special features against numerical unstability;
1518          some problems may cause premature termination of the solving
1519          when the matrix A*D*A' becomes ill-conditioned;
1521        * it computes only values of primal (auxiliary and structural)
1522          variables and doesn't compute values of dual variables (i.e.
1523          reduced costs) which are just set to zero;
1525        * it doesn't identify optimal basis corresponding to the found
1526          interior point solution; all variables in the found solution
1527          are just marked as basic variables.
1529        GLPK also includes a stand-alone program 'glpipm' which is a
1530        demo based on the interior point method. It may be used in the
1531        same way as the program 'glpsol' that is based on the simplex
1532        method.
Note: See TracBrowser for help on using the repository browser.