Re: [abcusers] reusable parser - Why an Output Format?

[Phil Taylor]

On 27 Apr 2004, at 18:37, Bernard Hill wrote:
In message <[EMAIL PROTECTED]>, Christian M. Cepel 
<[EMAIL PROTECTED]> writes
Exu Yangi wrote:
<Snip>
Ah, yes. What do we output? Once again, the number of output 
technologies available in common would seem to indicate either XML 
or INI format. They are text based, and portable assuming we ignore 
the wierd end-of-line marker problems on Macs.

I would vote for XML.
Maybe I'm stupid, but I don't understand this bit of the 
conversation. I would expect that a parser would not have ANY form of 
structured output for an entire tune, or an entire book of tunes.  I 
don't understand why anyone would want to write a parser to parse a 
file and then spit out a new file (or buffer) of stuff that then 
itself has to be re parsed to be used???
Simply because abc is such a mish-mash (read: "humanly readable") 
notation that the production of a structured output from it is the 
hard part. It's easy to write an XML parser (or other) in comparison.
It's true that it's much easier to parse XML than abc.  That doesn't 
make XML a good choice for the output of a general-purpose abc parser 
though.  What happens is that you convert an average tune of 250 bytes 
to 100K of XML, then parse all that huge volume of text again to get 
what you actually want.  That's going to be incredibly inefficient.

Actually, I don't think it's possible to make a single abc parser to 
suit all purposes.  In BarFly I ended up writing two separate parsers.  
One parses the tune for playing, generating midi and doing analysis;  
it's quite slow, very tolerant of syntax errors (reports them 
discreetly but mostly carries on anyway) and knows about stress 
programming.  The other parses the tune to produce a picture of the 
staff notation.  It's very fast (so it can be run after every keystroke 
and the screen display kept up to date during editing), much more 
critical of syntax errors, and it keeps track of correspondence between 
the abc symbols and the position of the corresponding notes in the 
picture, so that the user can select the text by clicking on the 
picture.

If you can come up with a single parser that will do all of that then I 
may use it.  Otherwise I'll just carry on with what I have.

Phil Taylor
To subscribe/unsubscribe, point your browser to: http://www.tullochgorm.com/lists.html