Re: strsplit needs an update

[Ben Abbott]

On Mar 27, 2013, at 12:50 PM, Jordi Gutiérrez Hermoso wrote:

> So this is why it's always an annoyance to try to implement something
> before Matlab... We've had strsplit since 2009 in Octave, and in 2013,
> TMW finally implements it themselves, with a differing calling form
> than ours:
> 
>    http://www.mathworks.com/help/matlab/ref/strsplit.html
> 
> Anyone feel like chasing after this?
> 
> - Jordi G. H.

I've prepared a version that is compatible with Matlab, but compatibility means 
that the current behavior must be broken.  There are many files which rely upon 
strsplit().

scripts/deprecated/javafields.m
scripts/deprecated/javamethods.m
scripts/general/fieldnames.m
scripts/general/int2str.m
scripts/general/methods.m
scripts/general/num2str.m
scripts/help/gen_doc_cache.m
scripts/help/help.m
scripts/help/lookfor.m
scripts/io/strread.m
scripts/java/javaclasspath.m
scripts/miscellaneous/compare_versions.m
scripts/miscellaneous/computer.m
scripts/miscellaneous/fact.m
scripts/miscellaneous/tar.m
scripts/miscellaneous/unpack.m
scripts/miscellaneous/what.m
scripts/miscellaneous/zip.m
scripts/pkg/private/configure_make.m
scripts/pkg/private/fix_depends.m
scripts/pkg/private/generate_lookfor_cache.m
scripts/pkg/private/list_forge_packages.m
scripts/pkg/private/unload_packages.m
scripts/pkg/private/write_index.m
scripts/plot/private/__file_filter__.m
scripts/plot/private/__fltk_file_filter__.m
scripts/plot/private/__go_draw_axes__.m
scripts/plot/private/__next_line_style__.m
scripts/strings/strsplit.m
scripts/strings/untabify.m
scripts/testfun/rundemos.m
scripts/testfun/runtests.m

I've modified the files to use the new version, and produced a changeset.  I 
don't see any regressions, but this may break some code on Octave-Forge.

The threee major impacts are;

(1) Delimiters are now string vectors, not scalars.

Old behavior

strsplit ("1 2, 3", ", ")
ans = 
{
  [1,1] = 1
  [1,2] = 2
  [1,3] = 
  [1,4] = 3
}

New (compatible) behavior

strsplit ("1 2, 3", ", ")
ans = 
{
  [1,1] = 1 2
  [1,2] = 3
}

(2) The Matlab default is to treat consecutive delimiters are treated as a 
single delimiter.  The old Octave default was the opposite.

(3) The old behavior supporting 2D character arrays is gone (only vectors now).

Comments?

Ben

changeset.patch
Description: Binary data