Re: Emacs as a translator's tool

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Jean-Christophe Helary <jean.christophe.helary@traduction-libre.org>
writes:

>> On May 30, 2020, at 3:22, Eric Abrahamsen <eric@ericabrahamsen.net> wrote:
>> 
>>>> I
>>>>  imagined this would happen when the mode was turned on: it would run
>>>>  down the file and insert markers that would be used to find
>>>>  correspondences. Special characters could be inserted into the file
>>>>  to indicate that two paragraphs should be joined, or one paragraph
>>>>  split.
>>> 
>>> What would be the use of such a marking ?

[...]

>> I want to be able to go to the
>> bottom of the translation, run a command, and have the second window
>> display the corresponding original. If I realize I've done something
>> wrong a couple of chapters previously, and I skip back up to that
>> location in the translation, I want to run the same command to display
>> the corresponding spot in the original.
>
> I seem to remember a long discussion about bookmarks here or on devel
> a while ago. Did you consider that ?

The only part of this code I ever actually wrote used bookmarks to save
where I was at the end of the work day. But usually you just save one
bookmark per file, indicating "where you are" in the file. That's a
different concern than splitting the two texts into segments, and
recording correspondences between segments in the texts. If you
segmented a whole novel by sentences, and then saved a bookmark per
sentence, I'm sure it would cause something to catch on fire.

At first I thought I'd run through the text when the mode was turned on,
insert a whole bunch of markers, then keep a list of marker-pairs. That
seemed like it would be hard to keep properly in sync, though, so now
I'm thinking of running through the text and actually inserting
separator characters, perhaps #x1f, either making them invisible or
putting some other nice display on them. That makes it easier to sync,
and has the advantage that it persists to disk and you only have to do
the major parsing once. Then strip them out during export.

Anyway, still experimenting...