Re: new strsplit function

[Philip Nienhuis]

Ben Abbott wrote:
> On Apr 20, 2013, at 3:51 PM, Philip Nienhuis wrote:
>
> > Ben Abbott wrote:
> > > > On Apr 2, 2013, at 11:42 PM, John W. Eaton wrote:
:
<snip>
:
> > > > > Also, the new strsplit does not seem to be working correctly for single-quoted 
> > > > > strings that contain escape sequences.  I think that
> > > > >
> > > > > strsplit ("foo\tbar", '\t')
> > > > >
> > > > > should split on the TAB, but it is currently returning the original string for 
> > > > > me.
> > > >
> > > > Opps.  I hadn't considered single quoted strings.  Your example does work in 
> > > > Matlab.  How is this sort of thing handled in other places?  It is sufficient 
> > > > to just ...

<snip>

> > And then there is the bug John mentioned with escape characters.
> > On a stable MinGW 3.6.4 build it doesn't work either:
> >
> > octave:1>  strsplit ("foo\tbar", '\t')
> > ans =
> > {
> >   [1,1] = foo   bar
> > }
>
> Is this with the changeset I attached applied?

Sorry I got a but confused, didn't pick up that your "sq" cs was for 
that very issue.

I applied it manually to yesterday's MXE-build (from yesterday's updated 
source tree) and now it works correctly:

octave-cli:1> test strsplit
PASSES 38 out of 38 tests
octave-cli:2> strsplit ("foo\tbar", '\t')
ans =
{
  [1,1] = foo
  [1,2] = bar
}
octave-cli:3>


Anyway, in view of JWE's comments, what is going to happen now with 
strsplit.m?
Do we keep the changes and the "legacy" option, or do we simply follow 
ML as John suggested?

As long as speed isn't affected (-significantly) for simple cases I 
don't care very much.
You have already patched the affected core functions.
In the io package (spreadsheet functions) there are only "simple" calls 
to strsplit (i.e. with just a one-char delimiter) so am I right to 
assume no changes are required there?

BTW in the texinfo help of strsplit there's neither an (explicit) 
explanation of what the 'simple' DELIMITERTYPE does, nor an example.

Philip