Re: [ELPA] New package proposal: org-translate

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Stefan Kangas <stefankangas@gmail.com> writes:

> Hi Eric,
>
> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> I've added version 0 of a package called org-translate to ELPA, and I'm
>> proposing it for release. It's a translation environment based on Org.
>> It's not quite a full-on CAT tool, but I suppose has the potential to
>> get there. I'm posting the Commentary section from the library below.
>> Comments/suggestions/complaints very welcome!
>
> I am happy to hear a package like this is in the works, as I do and have
> done a lot of translating in Emacs.
>
> I realize this is just a first draft, but I'll just add my thoughts
> from trying the package.

Thank you very much for testing!

> First, I think the Commentary should clearly explain how to set up your
> org file and get started.  It explains it only abstractly, but
> preferably should do so in concrete steps.
>
> The minor mode documentation should include an overview of the key
> bindings.  I would also appreciate an overview of the keybindings and
> expected workflow in the Commentary section.

Yep, others have also mentioned that the docs are lacking, I'll add this
first.

> `ogt-start-translating' while in an org buffer should just start with
> the current org buffer.  It works after setting up a bookmark, but I'm
> not sure requiring the user to keep a bookmark is the best design
> choice.  (Also, there is no completion for the bookmarks.)

`ogt-start-translating' is really only useful if you're not in the org
buffer already, and if you've chosen to use bookmarks. Maybe I should
rename it `ogt-open-bookmark' or something. Otherwise, just turning on
the minor mode is all you need.

You only get completion if you've already bookmarked a translation
project -- `start-translating' will only offer bookmarks that represent
existing translation projects.

> So I set up the bookmark, and now it asks to segment the file. Seems to
> work fine, and I get markers in my file (but all the markers say ""
> for some reason).  Now my file looks like:
>
>     * Original
>
>      In the early hours of 9 September
>
> The first segment is from after the "l" in "Original" to before "In".
> I guess that should rather be ignored.
>
> If I remove the headline and run `ogt-forward-segment', I see the
> following backtrace:

I think all this is just a symptom of the mode being unable to identify
which subtree is the original text, and which is the translation. I
think in that case it should just error out with a helpful message,
rather than limping along and segmenting incorrectly.

[...]

> Saving the file and opening it again, I now see " " in every location
> where there previously were markers. Running `ogt-start-translating'
> doesn't give a backtrace, but the message: "org-back-to-heading: Before
> first headline at position 121 in buffer trans.org"
>
> So I re-add the "Original" headline and try again, seems to work now and
> the characters above are again the markers "". I'm not sure if I'm
> supposed to type the text directly in place? I would rather have the
> translation text in a separate window, fixed in place and highlighted,
> and I'd like to see only the segment I'm currently translating with the
> rest of the text highlighted. More like OmegaT or a proper CAT
> software.

I'm still thinking this is a result of the mode misidentifying the parts
of the buffer. Your notes above have just " " and "" where you're
talking about the markers -- I'm assuming the actual characters were
stripped out on sending.

If you get the parts properly identified, I think the mode will behave
the way you want it to: source text in one window, translation in
another, and the source window follows along as you progress with the
translation.

> That's basically as far as I got for now.

Thanks for test-running what's obviously a half-baked library! I guess I
hadn't yet gotten out of my own head and thought about how a new user
would approach the code.

Step one should be making sure that the buffer contains one top-level
heading for each of the source text, translation text, and glossary.
Look at the values (and docstrings) of the
`ogt-default-{source,translation,glossary}-locator' options to see how
to tell the minor mode to locate each subtree.

But I'll put all this in the Commentary section later today.

Thanks again,
Eric