[TxMt] REQUEST: Scope-based Find/Replace

[Alain O'Dea]

On 30-Mar-08, at 3:18 PM, Thomas Aylott - subtleGradient wrote:

> On Mar 29, 2008, at 10:37 PM, Alain O'Dea wrote:
>> I would find it incredibly useful if I could do a Find/Replace such  
>> that I could reliably rename symbols like variables and function  
>> names. I envision that this would be based on scopes and that it  
>> would constrain Find/Replace to exact matches of entire scopes. In  
>> Erlang this would enable me to rename variables by doing a Find/ 
>> Replace of exact scope match for variable.other.erlang.
>
> Actually, search and replace is more powerful than the scope  
> matching in language grammars.
> 	Find and replace regex can match multiple lines.
>
> What you could do is look at the language definition file and find  
> out the regex that's used to scope variables and then modify it to  
> search and replace the specific variables you're looking for.
>
> Maybe this could be automated in a command somehow. But, currently  
> TextMate wouldn't be able to do what you're asking in any but the  
> currently open document, since it doesn't even parse any files that  
> it doesn't have open, and a command isn't capable of accessing the  
> parsed version of any but the frontmost document anyway.
>
> Maybe you could do something with a ruby script that parses all the  
> files in the project or something similar and asks you what  
> variables you want to replace.
>
> Personally, I wouldn't touch something like this because of the pain  
> that it would take to debug it.

I see how your suggestion would work for Erlang, where variables are  
lexically distinct from other symbols. However, just grabbing the  
regex for variables from the language file won't suffice because it  
loses the state/context in which that regex validly extracts variables  
from the stream of text. Even in Erlang, this is a problem because  
capitalized words (which Erlang's variables must be) can appear in  
comments, strings, and other contexts where they are not logically  
variables.

You mentioned by that a command isn't capable of accessing the parsed  
version of any of the frontmost document. Does that mean that the  
parsed version of the frontmost document is somehow accessible to  
commands? That would be immensely useful to me for a variety of  
purposes.