[Scilab-users] {EXT} how to faster EVSTR?

[Samuel Gougeon]

Le 21/10/2019 à 16:39, Dang Ngoc Chan, Christophe a écrit :
> Hello,
>
>> De : anna78
>> Envoyé : lundi 21 octobre 2019 16:02
>>
>> test_number=evstr(test_str)
>>
>> It looks like working but it takes a not acceptable time...
>>
>> Do you have a more clever idea?
> You might try strtod()

Right.
evstr() is slower mainly because, historically, it includes some 
concatenation features that enables it not working in an element-wise 
way, like with

--> evstr(["1:3" "4" ; "5" "7:9"])
  ans  =
    1.   2.   3.   4.
    5.   7.   8.   9.

This was an unexpected (and undocumented) complication when debugging 
and upgrading
evstr() 2 years ago. We had to (temporarily) renounce to some 
element-wise processing.

However, in order to avoid breaking the back-compatibility, such a 
element-wise processing
could be implemented through a"by-element" optional flag. Then, the 
sizes of the input and
output would match, and some related bugs like the 10196 
<http://bugzilla.scilab.org/show_bug.cgi?id=10196> could be fixed, at last.

Samuel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20191028/a883a532/attachment.htm>