Re: XML tools for Octave

[Alois Schloegl]

Andy,

thanks for your comments.

Andy Adler wrote:

> On Sat, 10 Jun 2006, Schloegl Alois wrote:
>
> > This works very well. Although the OCT version is not as fast as the 
> > MEX version, it works also for a case, were the MEX function causes a 
> > bad crash.
> >
> > Andy, what do you think about the solution?
> > It does not recognize the example 1) as an invalid XML. How big of a 
> > problem is this?
>
>
> Alois,
>
> It depends on what you want to do with it.
>
> I did a comparison of the various XML tools for available for matlab,
> and wasn't impressed with any of them; this one has at least the 
> advantage
> of being simple and robust.
>
> However, it has a few defects:
>
> 1. Parses non-XML files ( <b><i>tag</b></i> )
>
> 2. These files are identical in the infoset, but are
>     parsed differently
>
>      <XML> <tag/> </XML>             vs <XML> <tag></tag> </XML>
>      <XML> &quot; </XML>             vs <XML> " </XML>
>      <XML> <![CDATA[ text ]]> </XML> vs <XML> text </XML>
>
> 3. Doesn't manage xml comments (<!-- -->)

Comments are loaded and appear in the substructure  data  as strings.

> 4. Confused by processing instructions
>      <?xml version="1.0" encoding="UTF-8"?>
What do you mean by "confused"? So far I've not observed a problem with 
this.

> Are these a big issue for you? I would suspect that 3 and
> 4 are probably quite important.


No, not really. I do not care about comments. I'm a little bit concerned 
though that some non-XML data could cause some indefinite state.

Currently, the biggest issue for me is that the outcome needs some 
postprocessing.
Instead of
      X.data.='aECG'
      X.sub = {anystruct}

I'd like to see
      X.aECG = {anystruct}

> What features are you looking for in a XML parser?

Basically, I want to load large XML data sets, e.g. like this one
http://hci.tugraz.at/schloegl/biosig/xml/Exa01.xml
and extract the information which is needed for the common interface on 
biomedical data formats in the "BioSig for Octave and Matlab" toolbox.

Moreover, reading OpenOffice Spreadsheet files (which are just zipped 
XML) would be a nice and useful feature.

> --
> Andy Adler <address@hidden> 1(613)562-5800x6218

I've included a link at the wiki site.


> APPENDIX: My comments on XML tools:
>
> 1. XML4MAT
>    http://bioinformatics.org/project/?group_id=172
>    LICENCE: GPL V2
>
>    *.m code throughout
>    Uses regexps to parse XML.
>    Fails for many simple examples (ie <tag att="<<"/>)
>
> 2. XML Toolbox for Matlab
>    http://www.geodise.org/toolboxes/generic/xml_toolbox.htm
>    LICENCE: Custom (BSD with advertising)
>
>    Output format looks interesting
>    Appears to only offer p-code downloads and to link
>    to Java xml parsers
>
>
> 3. XML Tree
>    http://www.artefact.tk/software/matlab/xml/
>    LICENCE: GPL
>
>    Extensive use of Matlab classes
>
>    BUGS:
>    Doesn't record attributes
>    Misses mixed contents:
>     Misses <b to end:  <i> adsf <b/> asbf </i>
>     Misses asbf:       <i> adsf <b> </b> asbf </i>
>
>    Confused by Processing Instructions
>    Confused by Entities
>    Confused by CDATA
>
>    Crashes with namespaces
>
>    OK:
>       XML procesing instructions
>       XML comments
>       XML entities
>
> 4. XML Tools
>    
> http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=1742&objectType=FILE 
>
>    LICENCE: GPL
>
>    BUGS:
>      Parses non XML: <c t="C">F S</cd>
>                      <b><i>tag</b></i>
>
>      Infoset equivalent files aren't: <a></a> ~= <a/>
>
>      Doesn't parse entities
>      Doesn't manage comments
>      Doesn't manage CDATA
>      Confusted by <?xml ?>
>
>    OK:
>     Multiple attributes