off-topic: Conversion multi-line diff to single-line

off-topic: Conversion multi-line diff to single-line

[Werner LEMBERG]

Folks,


I have an off-topic question, and maybe you can help.  Is there a
program or script which converts the multiline chunks of diff (without
-u or -c) into single-line chunks?  This is, I want to have this

   XXX
   < foo1
   < foo2
   < foo3
   ---
   > bar1
   > bar2
   > bar3

converted into

   XXX
   < foo1
   ---
   > bar1
   YYY
   < foo2
   ---
   > bar2
   ZZZ
   > foo3
   ---
   < bar3

Such an output greatly simplifies comparison of large sorted word
lists with sparse or systematic insertions, deletions, or changes.


    Werner

Re: off-topic: Conversion multi-line diff to single-line

[Stefan Monnier]

> I have an off-topic question, and maybe you can help.  Is there a
> program or script which converts the multiline chunks of diff (without
> -u or -c) into single-line chunks?  This is, I want to have this

>    XXX
>    < foo1
>    < foo2
>    < foo3
>    ---
>> bar1
>> bar2
>> bar3

> converted into

>    XXX
>    < foo1
>    ---
>> bar1
>    YYY
>    < foo2
>    ---
>> bar2
>    ZZZ
>> foo3
>    ---
>    < bar3

> Such an output greatly simplifies comparison of large sorted word
> lists with sparse or systematic insertions, deletions, or changes.

Very good question.  I'd also like to see such a thing (tho on
the unified-diff format since that's the one I prefer).  I.e. turn

   --- .dir-locals.el	2011-04-24 18:34:57 +0000
   +++ .dir-locals.el	2011-05-27 23:05:19 +0000
   @@ -1,6 +1,6 @@
    ((nil . ((tab-width . 8)
   -         (sentence-end-double-space . t)
   -         (fill-column . 70)))
   +         (sentence-end-double-space . nil)
   +         (foo . 5)
   +         (fill-column . 79)
   +         (bar . 10)))
     (c-mode . ((c-file-style . "GNU")))

into

   --- .dir-locals.el	2011-04-24 18:34:57 +0000
   +++ .dir-locals.el	2011-05-27 23:05:19 +0000
   @@ -1,6 +1,6 @@
    ((nil . ((tab-width . 8)
   -         (sentence-end-double-space . t)
   +         (sentence-end-double-space . nil)
   +         (foo . 5)
   -         (fill-column . 70)))
   +         (fill-column . 79)
   +         (bar . 10)))
     (c-mode . ((c-file-style . "GNU")))

Your example is simpler but I think it's possible to handle mine
as well.  The needed info is available to diff-refine-hunk, so it should
be possible to extend diff-refine-hunk to diff-reorganize-hunk so as to
reduce the distance between matching words in the + and - lines.
Patches welcome.


        Stefan

Re: off-topic: Conversion multi-line diff to single-line

[Juri Linkov]

> Such an output greatly simplifies comparison of large sorted word
> lists with sparse or systematic insertions, deletions, or changes.

I suggest you using `diff --side-by-side --suppress-common-lines'
as is or by converting its output to the Diff normal format
that would be trivial to do from the side-by-side format.

Re: off-topic: Conversion multi-line diff to single-line

[Werner LEMBERG]

>> Such an output greatly simplifies comparison of large sorted word
>> lists with sparse or systematic insertions, deletions, or changes.
> 
> I suggest you using `diff --side-by-side --suppress-common-lines' as
> is or by converting its output to the Diff normal format that would
> be trivial to do from the side-by-side format.

Thanks for the idea, but this is exactly equivalent to a normal diff
output.  Note that you can get output like this

       > foo1
       > foo2
       > foo3
       > foo4
  bar1 <
  bar2 <
  bar3 <

which is no longer trivial to convert.


    Werner

Re: off-topic: Conversion multi-line diff to single-line

[Juri Linkov]

> Thanks for the idea, but this is exactly equivalent to a normal diff
> output.  Note that you can get output like this
>
>        > foo1
>        > foo2
>        > foo3
>        > foo4
>   bar1 <
>   bar2 <
>   bar3 <
>
> which is no longer trivial to convert.

I don't know what kind of input can produce such an output.
When I run `diff --side-by-side --suppress-common-lines' it returns:

  foo1 | bar1
  foo2 | bar2
  foo3 | bar3
  foo4 <

Writing a script that replaces " | " with "< " and " >" you can get:

  < foo1
  > bar1
  < foo2
  > bar2
  < foo3
  > bar3
  < foo4

that is not equivalent to a normal diff output that groups adjacent lines
into diff hunks like:

  < foo1
  < foo2
  < foo3
  < foo4
  > bar1
  > bar2
  > bar3

Re: off-topic: Conversion multi-line diff to single-line

[Werner LEMBERG]

>> Note that you can get output like this
>>
>>        > foo1
>>        > foo2
>>        > foo3
>>        > foo4
>>   bar1 <
>>   bar2 <
>>   bar3 <
>>
>> which is no longer trivial to convert.
> 
> I don't know what kind of input can produce such an output.

:-)  Here an example:

  Ab-na-be-lungs|pro-zeß   | Ab-na-be-lungs|pro-zess
  Ab·neh·mer·in.du·strie   | Ab·neh·mer·in.dus·trie
  Ab·neh·mer·in.du·stri·en | Ab·neh·mer·in.dus·tri·en
                           > ab-ni-cken
                           > Ab-ni-ckens
                           > Ab-ni-cker
                           > Ab-ni-ckern
  ab·ni{ck/k·k}en          <
  Ab·ni{ck/k·k}ens         <
  Ab·ni{ck/k·k}er          <
  Ab·ni{ck/k·k}ern         <
  ab-nor-mal-ste           | ab-nor-mals-te   
  ab-nor-mal-stem          | ab-nor-mals-tem
  ab-nor-mal-sten          | ab-nor-mals-ten

I suppose that `diff' is doing this if the lines are too different,
according to a certain built-in algorithm.

BTW, I also have this:

  ab·pa{ck/k·k}en          | ab-pa-cken
  Ab·pa{ck/k·k}er          | Ab-pa-cker
  Ab·pa{ck/k·k}ung         <
  Ab·pa{ck/k·k}un·gen      <
                           > Ab-pa-ckung
                           > Ab-pa-ckun-gen
  ab-paßt                  | ab-passt
  ab-paß-te                | ab-pass-te

(this is, first lines with < and then lines with >).


    Werner