Re: strsplit

[Ben Abbott]

On Jun 5, 2014, at 8:04 PM, Juan Pablo Carbajal <address@hidden> wrote:

> On Fri, Jun 6, 2014 at 2:01 AM, Mike Miller <address@hidden> wrote:
>> On Fri, Jun 6, 2014 at 01:05:53 +0200, Juan Pablo Carbajal wrote:
>>> Hi,
>>> 
>>> Is the following behavior correct in strsplit?
>>> 
>>> strsplit("/usr/local/share/octave/",filesep())
>>> ans =
>>> {
>>>  [1,1] =
>>>  [1,2] = usr
>>>  [1,3] = local
>>>  [1,4] = share
>>>  [1,5] = octave
>>>  [1,6] =
>>> }
>>> 
>>> strsplit("/usr/local/share/octave/",filesep(),true)
>>> ans =
>>> {
>>>  [1,1] =
>>>  [1,2] = usr
>>>  [1,3] = local
>>>  [1,4] = share
>>>  [1,5] = octave
>>>  [1,6] =
>>> }
>>> 
>>> strsplit("/usr/local/share/octave/",filesep(),false)
>>> ans =
>>> {
>>>  [1,1] =
>>>  [1,2] = usr
>>>  [1,3] = local
>>>  [1,4] = share
>>>  [1,5] = octave
>>>  [1,6] =
>>> }
>>> 
>>> 
>>> The output puts an empty cell element when the matching delimiter is
>>> at the beginning and at the end.
>>> This causes geometry ADD_PKG script to fail, so the behavior changed
>>> not long ago (I remember some mails about this, ml compatibility?).
>> 
>> I see the same behavior in 3.4.3, 3.6.4, and 3.8.1 for this example.
>> See also the ostrsplit function in 3.8 which operates like the old
>> strsplit.
>> 
>> Anyway, according to bug #39010 [1], it looks like yes, a string that
>> begins and ends with the delimiter should return an empty string from
>> both ends.
>> 
>> [1] https://savannah.gnu.org/bugs/?39010
>> 
>> --
>> mike
> 
> Thank you fro the quick response and the tests.
> 
> ok, I guess this is hereditary disease (or can anybody justify this 
> behavior?).
> If the behavior is so old I wonder how I wasn't getting the failure before.
> Anyways I replaced the use of strsplit with fullfile, hopefully this
> is more robust and portable.

The commands in Matlab would be ...

        strsplit('/usr/local/share/octave/',filesep())
        strsplit('/usr/local/share/octave/',filesep(),'CollapseDelimiters',true)
        
strsplit('/usr/local/share/octave/',filesep(),'CollapseDelimiters',false)

The third arg for Octave is being taken to be equivalent to Matlab's 
"CollapseDelimiters" parameter.

Perhaps Octave's third arg can be made equivalent to "IgnoreEmpties" instead of 
"CollapseDelimiters"?

In any event, the old version of strsplit() is still present, but was renamed.

        ostrsplit('/usr/local/share/octave/',filesep(),true)
        ans =
        {
                [1,1] = usr
                [1,2] = local
                [1,3] = share
                [1,4] = octave
        } 

Ben