Re: Idea: How about logical lines?

[Per Dalgas Jakobsen <pdj knaldgas dk>]

On 11/05/2016 05:42 AM, andré wrote:
> Per Dalgas Jakobsen a écrit :
> > Now I'm looking for a way to let Meld ignore formatting of the code,
> > ...
> > Is it a good idea to let Meld (and other diffs) find changes based on
> > logical lines as an alternative to physical lines? Like a
> > switch/preference to use logical lines instead (with options for
> > line-continuation and line-end).
> > Have I overlooked some important considerations here?
>
> Interesting idea, but wouldn't meld have to understand the structure 
> of the language coded ?
>
> It seems to me that it would be difficult to code such comparisons, 
> but I could be mistaken.  Breaking out the logical lines would be one 
> step, but for the comparison to be useful, it would also have to 
> locate the logical lines in the initial (physical line) file.
>
> If one is comparing different changes to the same file, I wouldn't 
> expect that a logical line comparison would produce vastly different 
> results than a physical line comparison. 

Well, Meld does not have to understand the code structure as such, only 
how a logical line is defined (for the language in question). I believe 
that having Meld ignoring comments by regex-filtering, the same could be 
done for logical lines.

First, if one is interested in comparing source code functionality, one 
is almost certainly not interested in changes in whitespaces or 
comments; so comments could be filtered out first to simplify the following.
An regex could define a "line-continuation", e.g. "\\ *$" (backslash at 
the end of a line, with optional spaces in between). Logical line ending 
could be defined as "; *$" (semicolon at the end of a line, with 
optional spaces in between).

Then in the internal search, line-continuation is just a whitespace, 
CR/LF is whitespace unless it matches the Logical line ending.

I think these two regex will allow most languages to be diffed on 
logical lines (ignoring code formatting).
1) Languages which have logical lines ending at CR/LF, but can break 
lines with continuation character will use: Line_continuation="\\ *$" 
and Logical_EOL="$"
2) Languages which have no line-continuation concept (all whitespaces 
can break), will use: Line_continuation="" and Logical_EOL="; *$"
3) Languages with both, can use both :)

I know that having logical lines ending with a non CR/LF character 
allows for multiple logical lines in one physical line, and if Melds 
internal diffing is based on physical lines (GUI and all) which I 
suspect, it wont be able to ignore formatting where multiple logical 
lines is combined/split into one/multiple physical lines, but I don't 
think that would be a big issue in most cases.

~Per