max does not support sparse boolean matrix

[pbellec]

Bug report for Octave 3.0.4 configured for i686-pc-linux-gnu

Description:
-----------

Running MAX on a sparse boolean matrix results in an error message. As MAX supports both boolean and sparse matrix, I suspect this type was just omitted by the developer and that may qualify for being a "bug". I looked in the forum for "max sparse boolean" but I could not find a relevant post. The matlab version of MAX supports sparse boolean matrices.

Repeat-By:
---------

octave-3.0.4:7> a = sparse([true false])
a =

Compressed Column Sparse (rows = 1, cols = 2, nnz = 1)

  (1, 1) ->  1

octave-3.0.4:8> max(a)
error: max: wrong type argument `sparse bool matrix'
octave-3.0.4:8> max(double(a))
ans = Compressed Column Sparse (rows = 1, cols = 1, nnz = 1)

  (1, 1) ->  1
octave-3.0.4:9> max(full(a))
ans =  1

Fix:
---
MAX is a built-in variable and my knowledge in C is limited so I didn't look into the octave source code.

At the .m level, a workaround would be to implement the following if a is a sparse boolean :
max_a = max(double(a))>0
The output is of the right type.

Configuration (please do not edit this section):
-----------------------------------------------

uname output:     Linux sorbier 2.6.24-24-generic #1 SMP Wed Mar 25 12:16:54 UTC 2009 i686 GNU/Linux
configure opts:  
Fortran compiler: g77
FFLAGS:           -O -mieee-fp
F2C:              @F2C@
F2CFLAGS:         @F2CFLAGS@
FLIBS:            -L/usr/lib/gcc/i486-linux-gnu/3.4.6 -L/usr/lib/gcc/i486-linux-gnu/3.4.6/../../../../lib -L/usr/lib/gcc/i486-linux-gnu/3.4.6/../../.. -L/lib/../lib -L/usr/lib/../lib -lhdf5 -lz -lfrtbegin -lg2c -lm
CPPFLAGS:        
INCFLAGS:         -I. -I. -I./liboctave -I./src -I./libcruft/misc
C compiler:       gcc, version 4.2.4 (Ubuntu 4.2.4-1ubuntu4)
CFLAGS:           -g -O2
CPICFLAG:         -fPIC
C++ compiler:     g++, version 4.2.4
CXXFLAGS:         -g -O2
CXXPICFLAG:       -fPIC
LD_CXX:           g++
LDFLAGS:          
LIBFLAGS:         -L.
RLD_FLAG:         -Wl,-rpath -Wl,/usr/local/lib/octave-3.0.4
BLAS_LIBS:        -llapack -lblas
FFTW_LIBS:        -lfftw3
LIBS:             -lreadline  -lncurses -ldl -lhdf5 -lz -lm
LEXLIB:          
LIBGLOB:          
SED:              /bin/sed
DEFS:

  -DPACKAGE_NAME="" -DPACKAGE_TARNAME="" -DPACKAGE_VERSION=""
  -DPACKAGE_STRING="" -DPACKAGE_BUGREPORT="" -DOCTAVE_SOURCE=1
  -D_GNU_SOURCE=1 -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1
  -DHAVE_SYS_STAT_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1
  -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 -DHAVE_INTTYPES_H=1 -DHAVE_STDINT_H=1
  -DHAVE_UNISTD_H=1 -DSEPCHAR=':' -DSEPCHAR_STR=":" -D__NO_MATH_INLINES=1
  -DCXX_NEW_FRIEND_TEMPLATE_DECL=1 -DCXX_ISO_COMPLIANT_LIBRARY=1
  -DCXX_ABI=gnu_v3 -DHAVE_LIBM=1 -DHAVE_QHULL=1 -DHAVE_PCRE=1
  -DHAVE_REGEXEC=1 -DHAVE_REGEX=1 -DHAVE_ZLIB_H=1 -DHAVE_ZLIB=1
  -DHAVE_HDF5_H=1 -DHAVE_HDF5=1 -DHAVE_H5GGET_NUM_OBJS=1 -DHAVE_FFTW3=1
  -DHAVE_GLPK_H=1 -DHAVE_GLPK=1 -DHAVE_IEEE754_DATA_FORMAT=1
  -DF77_FUNC(name,NAME)=name ## _ -DF77_FUNC_(name,NAME)=name ## __
  -DHAVE_BLAS=1 -DHAVE_SUITESPARSE_UMFPACK_H=1 -DHAVE_UMFPACK=1
  -DUMFPACK_SEPARATE_SPLIT=1 -DHAVE_SUITESPARSE_COLAMD_H=1
  -DHAVE_COLAMD=1 -DHAVE_SUITESPARSE_CCOLAMD_H=1 -DHAVE_CCOLAMD=1
  -DHAVE_SUITESPARSE_CHOLMOD_H=1 -DHAVE_CHOLMOD=1 -DHAVE_SUITESPARSE_CS_H=1
  -DHAVE_CXSPARSE=1 -DHAVE_GETHOSTNAME=1 -DHAVE_GETPWNAM=1 -DHAVE_DEV_T=1
  -DHAVE_INO_T=1 -DHAVE_NLINK_T=1 -DHAVE_NLINK_T=1 -DHAVE_LONG_LONG_INT=1
  -DHAVE_UNSIGNED_LONG_LONG_INT=1 -DHAVE_SIGSET_T=1 -DHAVE_SIG_ATOMIC_T=1
  -DSIZEOF_SHORT=2 -DSIZEOF_INT=4 -DSIZEOF_LONG=4 -DSIZEOF_LONG_LONG=8
  -DHAVE_ALLOCA_H=1 -DHAVE_ALLOCA=1 -DNPOS=std::string::npos
  -DHAVE_PLACEMENT_DELETE=1 -DHAVE_DYNAMIC_AUTO_ARRAYS=1 -DSTDC_HEADERS=1
  -DHAVE_DIRENT_H=1 -DTIME_WITH_SYS_TIME=1 -DHAVE_SYS_WAIT_H=1
  -DHAVE_ASSERT_H=1 -DHAVE_CURSES_H=1 -DHAVE_DLFCN_H=1 -DHAVE_FCNTL_H=1
  -DHAVE_FLOAT_H=1 -DHAVE_GRP_H=1 -DHAVE_INTTYPES_H=1 -DHAVE_LIMITS_H=1
  -DHAVE_LOCALE_H=1 -DHAVE_MEMORY_H=1 -DHAVE_NCURSES_H=1 -DHAVE_POLL_H=1
  -DHAVE_PWD_H=1 -DHAVE_STDINT_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1
  -DHAVE_SYS_IOCTL_H=1 -DHAVE_SYS_PARAM_H=1 -DHAVE_SYS_POLL_H=1
  -DHAVE_SYS_RESOURCE_H=1 -DHAVE_SYS_SELECT_H=1 -DHAVE_SYS_STAT_H=1
  -DHAVE_SYS_TIME_H=1 -DHAVE_SYS_TIMES_H=1 -DHAVE_SYS_TYPES_H=1
  -DHAVE_SYS_UTSNAME_H=1 -DHAVE_TERMCAP_H=1 -DHAVE_UNISTD_H=1
  -DHAVE_UTIME_H=1 -DHAVE_SSTREAM=1 -DHAVE_TERMIO_H=1 -DHAVE_SGTTY_H=1
  -DHAVE_GLOB_H=1 -DHAVE_FNMATCH_H=1 -DHAVE_FNMATCH=1 -DHAVE_GLOB=1
  -DHAVE_ATEXIT=1 -DHAVE_BASENAME=1 -DHAVE_BCOPY=1 -DHAVE_BZERO=1
  -DHAVE_CANONICALIZE_FILE_NAME=1 -DHAVE_CHMOD=1 -DHAVE_DUP2=1
  -DHAVE_ENDGRENT=1 -DHAVE_ENDPWENT=1 -DHAVE_EXECVP=1 -DHAVE_FCNTL=1
  -DHAVE_FORK=1 -DHAVE_GETCWD=1 -DHAVE_GETEGID=1 -DHAVE_GETEUID=1
  -DHAVE_GETGID=1 -DHAVE_GETGRENT=1 -DHAVE_GETGRGID=1 -DHAVE_GETGRNAM=1
  -DHAVE_GETPGRP=1 -DHAVE_GETPID=1 -DHAVE_GETPPID=1 -DHAVE_GETPWENT=1
  -DHAVE_GETPWUID=1 -DHAVE_GETTIMEOFDAY=1 -DHAVE_GETUID=1 -DHAVE_GETWD=1
  -DHAVE_KILL=1 -DHAVE_LGAMMA=1 -DHAVE_LINK=1 -DHAVE_LOCALTIME_R=1
  -DHAVE_LSTAT=1 -DHAVE_MEMMOVE=1 -DHAVE_MKDIR=1 -DHAVE_MKFIFO=1
  -DHAVE_MKSTEMP=1 -DHAVE_ON_EXIT=1 -DHAVE_PIPE=1 -DHAVE_POLL=1
  -DHAVE_PUTENV=1 -DHAVE_RAISE=1 -DHAVE_READLINK=1 -DHAVE_REALPATH=1
  -DHAVE_RENAME=1 -DHAVE_RINDEX=1 -DHAVE_RMDIR=1 -DHAVE_ROUND=1
  -DHAVE_SELECT=1 -DHAVE_SETGRENT=1 -DHAVE_SETLOCALE=1
  -DHAVE_SETPWENT=1 -DHAVE_SETVBUF=1 -DHAVE_SIGACTION=1
  -DHAVE_SIGLONGJMP=1 -DHAVE_SIGPENDING=1 -DHAVE_SIGPROCMASK=1
  -DHAVE_SIGSUSPEND=1 -DHAVE_SNPRINTF=1 -DHAVE_STAT=1 -DHAVE_STRCASECMP=1
  -DHAVE_STRDUP=1 -DHAVE_STRERROR=1 -DHAVE_STRNCASECMP=1 -DHAVE_STRPTIME=1
  -DHAVE_STRSIGNAL=1 -DHAVE_SYMLINK=1 -DHAVE_TEMPNAM=1 -DHAVE_TGAMMA=1
  -DHAVE_UMASK=1 -DHAVE_UNAME=1 -DHAVE_UNLINK=1 -DHAVE_USLEEP=1
  -DHAVE_UTIME=1 -DHAVE_VFPRINTF=1 -DHAVE_VSPRINTF=1 -DHAVE_VSNPRINTF=1
  -DHAVE_WAITPID=1 -DHAVE_STRFTIME=1 -DHAVE_LIBDL=1 -DHAVE_DLOPEN=1
  -DHAVE_DLSYM=1 -DHAVE_DLERROR=1 -DHAVE_DLCLOSE=1 -DHAVE_DLOPEN_API=1
  -DENABLE_DYNAMIC_LINKING=1 -DHAVE_TIMEVAL=1 -DHAVE_FINITE=1
  -DHAVE_ISNAN=1 -DHAVE_ISINF=1 -DHAVE_COPYSIGN=1 -DHAVE_DECL_SIGNBIT=1
  -DHAVE_ACOSH=1 -DHAVE_ASINH=1 -DHAVE_ATANH=1 -DHAVE_ERF=1 -DHAVE_ERFC=1
  -DHAVE_EXP2=1 -DHAVE_LOG2=1 -DHAVE_STRUCT_STAT_ST_BLKSIZE=1
  -DHAVE_STRUCT_STAT_ST_BLOCKS=1 -DHAVE_STRUCT_STAT_ST_RDEV=1
  -DHAVE_STRUCT_TM_TM_ZONE=1 -DHAVE_TM_ZONE=1 -DUSE_READLINE=1
  -DEXCEPTION_IN_MATH=1 -DRETSIGTYPE=void -DHAVE_DECL_SYS_SIGLIST=1
  -DHAVE_POSIX_SIGNALS=1 -DRETSIGTYPE_IS_VOID=1 -DHAVE_GETRUSAGE=1
  -DHAVE_TIMES=1 -DYYTEXT_POINTER=1 -DGNUPLOT_BINARY="gnuplot"

User-preferences (please do not edit this section):
--------------------------------------------------

  EDITOR = emacs
  EXEC_PATH = /usr/local/libexec/octave/3.0.4/site/exec/i686-pc-linux-gnu:/usr/local/libexec/octave/api-v32/site/exec/i686-pc-linux-gnu:/usr/local/libexec/octave/site/exec/i686-pc-linux-gnu:/usr/local/libexec/octave/3.0.4/exec/i686-pc-linux-gnu:/usr/local/bin:/home/pbellec/quarantines/Feb-14-2008/bin:/bin:/usr/bin:/sbin:/usr/sbin:/usr/local/X11/bin:/usr/bin/:/usr/local/bic/bin:/usr/local/bin
  IMAGE_PATH = .:/usr/local/share/octave/3.0.4/imagelib
  PAGER = less
  PS1 = \s:\#>
  PS2 = >
  PS4 = +
  beep_on_error = 0
  completion_append_char =  
  crash_dumps_octave_core = 1
  echo_executing_commands = 0
  fixed_point_format = 0
  gnuplot_binary = gnuplot
  gnuplot_command_end =

  gnuplot_command_plot = pl
  gnuplot_command_replot = rep
  gnuplot_command_splot = sp
  gnuplot_command_title = t
  gnuplot_command_using = u
  gnuplot_command_with = w
  history_file = /home/pbellec/.octave_hist
  history_size = 1024
  ignore_function_time_stamp = system
  info_file = /usr/local/share/info/octave.info
  info_program = info
  makeinfo_program = makeinfo
  max_recursion_depth = 256
  output_max_field_width = 5
  output_precision = 5
  page_output_immediately = 0
  page_screen_output = 1
  print_answer_id_name = 1
  print_empty_dimensions = 1
  save_precision = 16
  saving_history = 1
  sighup_dumps_octave_core = 1
  sigterm_dumps_octave_core = 1
  silent_functions = 0
  split_long_rows = 1
  string_fill_char =  
  struct_levels_to_print = 2
  suppress_verbose_help_message = 0

[dbateman2]

pbellec wrote:

> Bug report for Octave 3.0.4 configured for i686-pc-linux-gnu
>
> Description:
> -----------
>
> Running MAX on a sparse boolean matrix results in an error message. As MAX
> supports both boolean and sparse matrix, I suspect this type was just
> omitted by the developer and that may qualify for being a "bug". I looked in
> the forum for "max sparse boolean" but I could not find a relevant post. The
> matlab version of MAX supports sparse boolean matrices.
>
>  

Something like the attached should make the behavior the same as the

full types (what does matlab do in this case?). Though I can't push the
patch at the moment as I'm in the middle of a rebuild

D.

--
David Bateman                                dbateman@...
35 rue Gambetta                              [phone] (Home)
92100 Boulogne-Billancourt FRANCE            [phone] (Mob)



diff --git a/src/DLD-FUNCTIONS/max.cc b/src/DLD-FUNCTIONS/max.cc
--- a/src/DLD-FUNCTIONS/max.cc
+++ b/src/DLD-FUNCTIONS/max.cc
@@ -468,10 +468,9 @@
  \
   if (single_arg && (nargout == 1 || nargout == 0)) \
     { \
-      if (arg1.type_id () == octave_sparse_matrix::static_type_id ()) \
+      if (arg1.is_real_type ()) \
  retval(0) = arg1.sparse_matrix_value () .FCN (dim); \
-      else if (arg1.type_id () == \
-       octave_sparse_complex_matrix::static_type_id ()) \
+      else if (arg1.is_complex_type ()) \
  retval(0) = arg1.sparse_complex_matrix_value () .FCN (dim); \
       else \
  gripe_wrong_type_arg (#FCN, arg1); \
@@ -480,10 +479,9 @@
     { \
       Array2<octave_idx_type> index; \
  \
-      if (arg1.type_id () == octave_sparse_matrix::static_type_id ()) \
+      if (arg1.is_real_type ()) \
  retval(0) = arg1.sparse_matrix_value () .FCN (index, dim); \
-      else if (arg1.type_id () == \
-       octave_sparse_complex_matrix::static_type_id ()) \
+      else if (arg1.is_complex_type ()) \
  retval(0) = arg1.sparse_complex_matrix_value () .FCN (index, dim); \
       else \
  gripe_wrong_type_arg (#FCN, arg1); \


_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[pbellec]

Thanks for the quick answer David. I will try to apply the patch and recompile next week. I am not sure I correctly understands what the patch does though

David Bateman-2 wrote:

Something like the attached should make the behavior the same as the full types

If it is converting the max to a full array, that would not do. if max is just a scalar it would have no impact, but commands like max(a,[],1) can generate big vectors or arrays. It would be important to keep the sparse attribute, and also the boolean type for cases where memory load is a critical issue.

David Bateman-2 wrote:

(what does matlab do in this case?).

Matlab 7.4.0.336 (R2007a) is actually messing up the types. It keeps the array sparse, but goes from boolean to double :

>> a = sparse([true false; false false]);
>> whos a
  Name      Size            Bytes  Class      Attributes

  a         2x2                17  logical    sparse    
>> max_a = max(a,[],1)

max_a =

   (1,1)        1
>> whos max_a
  Name       Size            Bytes  Class     Attributes

  max_a      1x2                36  double    sparse    

I hope this helps,

Pierre

[Jaroslav Hajek-2]

On Fri, Apr 3, 2009 at 11:24 PM, pbellec <pbellec@...> wrote:

>
> Thanks for the quick answer David. I will try to apply the patch and
> recompile next week. I am not sure I correctly understands what the patch
> does though
>
> David Bateman-2 wrote:
>>
>> Something like the attached should make the behavior the same as the full
>> types
>>
>
> If it is converting the max to a full array, that would not do. if max is
> just a scalar it would have no impact, but commands like max(a,[],1) can
> generate big vectors or arrays. It would be important to keep the sparse
> attribute, and also the boolean type for cases where memory load is a
> critical issue.
>

No, it just converts a sparse bool matrix to sparse double. That's acceptable.
David, if you have limited access now, would you like me to apply your patch?

regards

--
RNDr. Jaroslav Hajek
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[dbateman2]

Jaroslav Hajek wrote:

> On Fri, Apr 3, 2009 at 11:24 PM, pbellec <pbellec@...> wrote:
>  
>> Thanks for the quick answer David. I will try to apply the patch and
>> recompile next week. I am not sure I correctly understands what the patch
>> does though
>>
>> David Bateman-2 wrote:
>>    
>>> Something like the attached should make the behavior the same as the full
>>> types
>>>
>>>      
>> If it is converting the max to a full array, that would not do. if max is
>> just a scalar it would have no impact, but commands like max(a,[],1) can
>> generate big vectors or arrays. It would be important to keep the sparse
>> attribute, and also the boolean type for cases where memory load is a
>> critical issue.
>>
>>    
>
> No, it just converts a sparse bool matrix to sparse double. That's acceptable.
> David, if you have limited access now, would you like me to apply your patch?
>
> regards
>
>  

its applied ;-)

D.


--
David Bateman                                dbateman@...
35 rue Gambetta                              [phone] (Home)
92100 Boulogne-Billancourt FRANCE            [phone] (Mob)

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave