Problem with regexp nested groups

Problem with regexp nested groups

[Marc Tfardy]

Hallo,

I have some problem with regexp and I hope someone could help me.


Assume we have following text in a buffer.

-- DATA ----------------------------------------------------------------
<DATA="some/file/sample1.mp3">
<DATA="some/file/sample2.mp3">
blablalba
<DATA="some/file/sample3.wav">
blabla
OBJ('some/file/sample4.mp3')
OBJ('some/file/sample5.au')
------------------------------------------------------------------------

My goal is to extract some text from the buffer, namely only
portion of text between `DATA="' and `"' or between `OBJ('' and
`''. In both cases text must end with `.mp3' and the left and
right delimeter shoud be ignored. So the right result should look
like this:

some/file/sample1.mp3
some/file/sample2.mp3
some/file/sample4.mp3


I wrote this test function:

(defun get-data ()
   (interactive)
   (if (re-search-forward 
"\\(DATA=\"\\(.*?\.mp3\\)\"\\|OBJ('\\(.*?\.mp3\\)')\\)" nil t)
       (message "found: %s" (match-string-no-properties 1))
     (message "failed")))

With (match-string-no-properties 1) I get this result:

found: DATA="some/file/sample1.mp3"
found: DATA="some/file/sample2.mp3"
found: OBJ('some/file/sample4.mp3')

Not good. Now I replace (match-string-no-properties 1) with
(match-string-no-properties 2) and I get only two results for
both DATA lines:

found: some/file/sample1.mp3
found: some/file/sample2.mp3
found: nil

The version with (match-string-no-properties 3) returns
only OBJ line:

found: some/file/sample4.mp3
found: nil


Now the final question: how can I get results like version 2 and
3 but at ones? I would be grateful for any help

regards

Marc

Re: Problem with regexp nested groups

[Lennart Borgman (gmail)]

Marc Tfardy wrote:
> Hallo,
> 
> I have some problem with regexp and I hope someone could help me.
> 
> 
> Assume we have following text in a buffer.
> 
> -- DATA ----------------------------------------------------------------
> <DATA="some/file/sample1.mp3">
> <DATA="some/file/sample2.mp3">
> blablalba
> <DATA="some/file/sample3.wav">
> blabla
> OBJ('some/file/sample4.mp3')
> OBJ('some/file/sample5.au')
> ------------------------------------------------------------------------
> 
> My goal is to extract some text from the buffer, namely only
> portion of text between `DATA="' and `"' or between `OBJ('' and
> `''. In both cases text must end with `.mp3' and the left and
> right delimeter shoud be ignored. So the right result should look
> like this:
> 
> some/file/sample1.mp3
> some/file/sample2.mp3
> some/file/sample4.mp3
> 
> 
> I wrote this test function:
> 
> (defun get-data ()
>   (interactive)
>   (if (re-search-forward 
> "\\(DATA=\"\\(.*?\.mp3\\)\"\\|OBJ('\\(.*?\.mp3\\)')\\)" nil t)
>       (message "found: %s" (match-string-no-properties 1))
>     (message "failed")))
> 
> With (match-string-no-properties 1) I get this result:
> 
> found: DATA="some/file/sample1.mp3"
> found: DATA="some/file/sample2.mp3"
> found: OBJ('some/file/sample4.mp3')
> 
> Not good. Now I replace (match-string-no-properties 1) with
> (match-string-no-properties 2) and I get only two results for
> both DATA lines:
> 
> found: some/file/sample1.mp3
> found: some/file/sample2.mp3
> found: nil
> 
> The version with (match-string-no-properties 3) returns
> only OBJ line:
> 
> found: some/file/sample4.mp3
> found: nil
> 
> 
> Now the final question: how can I get results like version 2 and
> 3 but at ones? I would be grateful for any help
> 
> regards
> 
> Marc

Could you perhaps look for both (match-string-no-properties 1) and dito 2?

Re: Problem with regexp nested groups

[Phil Carmody]

Marc Tfardy <m-t-o___CUT__IT@web.de> writes:
> Hallo,
>
> I have some problem with regexp and I hope someone could help me.
>
>
> Assume we have following text in a buffer.
>
> -- DATA ----------------------------------------------------------------
> <DATA="some/file/sample1.mp3">
> <DATA="some/file/sample2.mp3">
> blablalba
> <DATA="some/file/sample3.wav">
> blabla
> OBJ('some/file/sample4.mp3')
> OBJ('some/file/sample5.au')
> ------------------------------------------------------------------------
>
> My goal is to extract some text from the buffer, namely only
> portion of text between `DATA="' and `"' or between `OBJ('' and
> `''. In both cases text must end with `.mp3' and the left and
> right delimeter shoud be ignored. 
...
> "\\(DATA=\"\\(.*?\.mp3\\)\"\\|OBJ('\\(.*?\.mp3\\)')\\)" nil t)

How about grouping /DATA="/ or /OBJ('/ as match 1, then capture
the filename as 2, and end with /.mp3"/ or /.mp3'/ as match 3?
That would match malformed lines such as 
  
  DATA="foo.mp3'

but if that's a problem you can capture the quotes too and 
fix them in post-production.

Phil
-- 
Dear aunt, let's set so double the killer delete select all.
-- Microsoft voice recognition live demonstration

Re: Problem with regexp nested groups

[Marc Tfardy]

Lennart Borgman (gmail) schrieb:

> Could you perhaps look for both (match-string-no-properties 1) and dito 2?

This two cases was only example. In real application can occur clearly
more such patterns, so I looking for an universal solutuion.

reagrds

Marc

Re: Problem with regexp nested groups

[Marc Tfardy]

Phil Carmody schrieb:
> Marc Tfardy <m-t-o___CUT__IT@web.de> writes:
>> Hallo,
>>
>> I have some problem with regexp and I hope someone could help me.
>>
>>
>> Assume we have following text in a buffer.
>>
>> -- DATA ----------------------------------------------------------------
>> <DATA="some/file/sample1.mp3">
>> <DATA="some/file/sample2.mp3">
>> blablalba
>> <DATA="some/file/sample3.wav">
>> blabla
>> OBJ('some/file/sample4.mp3')
>> OBJ('some/file/sample5.au')
>> ------------------------------------------------------------------------
>>
>> My goal is to extract some text from the buffer, namely only
>> portion of text between `DATA="' and `"' or between `OBJ('' and
>> `''. In both cases text must end with `.mp3' and the left and
>> right delimeter shoud be ignored. 
> ...
>> "\\(DATA=\"\\(.*?\.mp3\\)\"\\|OBJ('\\(.*?\.mp3\\)')\\)" nil t)
> 
> How about grouping /DATA="/ or /OBJ('/ as match 1, then capture
> the filename as 2, and end with /.mp3"/ or /.mp3'/ as match 3?

Interesting idea and it really works and it is definitely better then
multiple (match-string-no-properties x). I've modified slightly
your suggestion and I put extension togother with filename:

"\\(DATA=\"\\|OBJ('\\)\\(.*?\.mp3\\)\\(\"\\|')\\)"

So I need only to refer (match-string-no-properties 2).


> That would match malformed lines such as 
>   
>   DATA="foo.mp3'
> 
> but if that's a problem you can capture the quotes too and 
> fix them in post-production.

This disadvantage I can get over. The really problem occur when
the extract part of text exist not at position 2 (in case of
another, more complicated regexp). In this case I have the same
type of problem like in my first posting.

regards

Marc

Re: Problem with regexp nested groups

[Johan Bockgård]

Marc Tfardy <m-t-o___CUT__IT@web.de> writes:

> Not good. Now I replace (match-string-no-properties 1) with
> (match-string-no-properties 2) and I get only two results for both
> DATA lines:
>
> found: some/file/sample1.mp3
> found: some/file/sample2.mp3
> found: nil

In CVS Emacs:

"\\(DATA=\"\\(.*?\.mp3\\)\"\\|OBJ('\\(?2:.*?\.mp3\\)')\\)"

** The regexp form \(?<num>:<regexp>\) specifies the group number
   explicitly.

-- 
Johan Bockgård

Re: Problem with regexp nested groups

[Marc Tfardy]

Johan Bockgård schrieb:
> Marc Tfardy <m-t-o___CUT__IT@web.de> writes:
> 
>> Not good. Now I replace (match-string-no-properties 1) with
>> (match-string-no-properties 2) and I get only two results for both
>> DATA lines:
>>
>> found: some/file/sample1.mp3
>> found: some/file/sample2.mp3
>> found: nil
> 
> In CVS Emacs:
> 
> "\\(DATA=\"\\(.*?\.mp3\\)\"\\|OBJ('\\(?2:.*?\.mp3\\)')\\)"
> 
> ** The regexp form \(?<num>:<regexp>\) specifies the group number
>    explicitly.

Great! One reason more to install CVS version :) Thanks.

regards

Marc