[git for translators] How to always generate conflicts for merges?

[git for translators] How to always generate conflicts for merges?

[Ilya Kantor]

Hi,

We're using Git to manage translations of an open-source book, and most of time it works well. But there's also a problem.

When we pull changes from upstream (English) to translation (e.g. Japanese), git auto-merges them.

Sometimes there conflicts, but not all the time.

For example, when a new file is added to English, it just gets auto-merged into Japanese.
But all new changes must be human-controlled, translated. 

Is there a way to force git always generate a conflict, even if changes could be auto-merged?

----
Ilya
https://javascript.info

Re: [git for translators] How to always generate conflicts for merges?

[Santiago Torres Arias]

[-- Attachment #1: Type: text/plain, Size: 1039 bytes --]

Hi,

I'm not completely sure if this is the best way to achieve this, but I
you could instruct a merge driver to mark all new files as unset.

Cheers!
-Santiago.

On Thu, Jul 25, 2019 at 04:42:48PM +0300, Ilya Kantor wrote:
> Hi,
> 
> We're using Git to manage translations of an open-source book, and most of time it works well. But there's also a problem.
> 
> When we pull changes from upstream (English) to translation (e.g. Japanese), git auto-merges them.
> 
> Sometimes there conflicts, but not all the time.
> 
> For example, when a new file is added to English, it just gets auto-merged into Japanese.
> But all new changes must be human-controlled, translated. 
> 
> Is there a way to force git always generate a conflict, even if changes could be auto-merged?
> 
> ----
> Ilya
> https://urldefense.proofpoint.com/v2/url?u=https-3A__javascript.info&d=DwIFAg&c=slrrB7dE8n7gBJbeO0g-IQ&r=yZMPY-APGKyVIX7HgQFZJA&m=fywBEjNJhlI37lKNhH6u90hEOBCF5PqmRunhDCRO0Fg&s=We1ibeYQn4Z4Gf53M4CUyENq_BHhavulOg20KJIRxOg&e= 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [git for translators] How to always generate conflicts for merges?

[Junio C Hamano]

Ilya Kantor <iliakan@gmail.com> writes:

> We're using Git to manage translations of an open-source book, and
> most of time it works well. But there's also a problem.
>
> When we pull changes from upstream (English) to translation
> (e.g. Japanese), git auto-merges them.
>
> Sometimes there conflicts, but not all the time.
>
> For example, when a new file is added to English, it just gets
> auto-merged into Japanese.  But all new changes must be
> human-controlled, translated.
>
> Is there a way to force git always generate a conflict, even if
> changes could be auto-merged?

I am not sure what the workflow here and if it makes sense.  When
you have a file, "chapter47.txt", whose original is in English, the
translation projects (there are N of them, one for each language)
will have their own "chapter47.txt" that has translated text in the
same place?  It looks to me that, working that way, the project for
translating into e.g. Japanese have no way to keep any of the
original English version, in which case why are you even "merging"
the English version in the first place?

I would have understood if the original "chapter47.txt" is translated
into "chapter47_ja.txt" and friends, like "chapter47_fr.txt", all of
which sit next to the original "chapter47.txt".  Then merging an
updated version of the original from time to time would make perfect
sense to me---that would give you a way to see what changed in the
original (e.g. "git show --first-parent -m -p master chapter47.txt")
to guide you find the places you would need to make corresponding
changes to the variant of your language, e.g. "chapter47_??.txt".

Re: [git for translators] How to always generate conflicts for merges?

[Ilya Kantor]

Hi Junio,

There's a repo for each language, with the same file structure.

For example, English version (upstream):
https://github.com/javascript-tutorial/en.javascript.info/blob/master/1-js/01-getting-started/1-intro/article.md

Japanese:
https://github.com/javascript-tutorial/ja.javascript.info/blob/master/1-js/01-getting-started/1-intro/article.md

As English version is updated, changes need to be delivered to translations.
That's done with "git pull upstream master" from translations.

As the text structure (paragraphs) is the same, usually merges give conflicts exactly in the places where English version changed.

Sometimes though, e.g. when a new chapter is added to upstream, the merge just goes through "successfully".

That's what I'd like to avoid, as all changes need to be human-controlled.

---
Ilya Kantor
https://javascript.info

> On 25 Jul 2019, at 19:37, Junio C Hamano <gitster@pobox.com> wrote:
> 
> Ilya Kantor <iliakan@gmail.com> writes:
> 
>> We're using Git to manage translations of an open-source book, and
>> most of time it works well. But there's also a problem.
>> 
>> When we pull changes from upstream (English) to translation
>> (e.g. Japanese), git auto-merges them.
>> 
>> Sometimes there conflicts, but not all the time.
>> 
>> For example, when a new file is added to English, it just gets
>> auto-merged into Japanese.  But all new changes must be
>> human-controlled, translated.
>> 
>> Is there a way to force git always generate a conflict, even if
>> changes could be auto-merged?
> 
> I am not sure what the workflow here and if it makes sense.  When
> you have a file, "chapter47.txt", whose original is in English, the
> translation projects (there are N of them, one for each language)
> will have their own "chapter47.txt" that has translated text in the
> same place?  It looks to me that, working that way, the project for
> translating into e.g. Japanese have no way to keep any of the
> original English version, in which case why are you even "merging"
> the English version in the first place?
> 
> I would have understood if the original "chapter47.txt" is translated
> into "chapter47_ja.txt" and friends, like "chapter47_fr.txt", all of
> which sit next to the original "chapter47.txt".  Then merging an
> updated version of the original from time to time would make perfect
> sense to me---that would give you a way to see what changed in the
> original (e.g. "git show --first-parent -m -p master chapter47.txt")
> to guide you find the places you would need to make corresponding
> changes to the variant of your language, e.g. "chapter47_??.txt".
> 

Re: [git for translators] How to always generate conflicts for merges?

[Sergey Lukashev]

Hi everyone!

As I see it, in Junio's approach you more likely to have a translation and original out of sync because you have to figure out whether the changes in the original are actually reflected in the translation, which could be a tedious thing to do again and again.

In the approach that Ilya uses you HAVE TO manually fix just what has changed: you will have a content in the original language in places not yet translated. This approach abuses git a bit and, as I understand, works because all the repos with the translations were probably forked from the original repo so they are NOT unrelated.

As the answer to the original problem I would ask if it really matters whether you have a conflict or not. Just fix conflicts, commit and review all the new stuff that was added in the commits just merged and fix it too. The process is still human-controlled: if you do forget to translate something, it will be left untraslated and is likely to be spotted soon. You want to have a conflict where there ain't one (a new content that does not exist in your translation). A conflict is when something has changed in different wayS since a common ancestor.


25.07.2019, 20:42, "Ilya Kantor" <iliakan@gmail.com>:
> Hi Junio,
>
> There's a repo for each language, with the same file structure.
>
> For example, English version (upstream):
> https://github.com/javascript-tutorial/en.javascript.info/blob/master/1-js/01-getting-started/1-intro/article.md
>
> Japanese:
> https://github.com/javascript-tutorial/ja.javascript.info/blob/master/1-js/01-getting-started/1-intro/article.md
>
> As English version is updated, changes need to be delivered to translations.
> That's done with "git pull upstream master" from translations.
>
> As the text structure (paragraphs) is the same, usually merges give conflicts exactly in the places where English version changed.
>
> Sometimes though, e.g. when a new chapter is added to upstream, the merge just goes through "successfully".
>
> That's what I'd like to avoid, as all changes need to be human-controlled.
>
> ---
> Ilya Kantor
> https://javascript.info
>
>>  On 25 Jul 2019, at 19:37, Junio C Hamano <gitster@pobox.com> wrote:
>>
>>  Ilya Kantor <iliakan@gmail.com> writes:
>>
>>>  We're using Git to manage translations of an open-source book, and
>>>  most of time it works well. But there's also a problem.
>>>
>>>  When we pull changes from upstream (English) to translation
>>>  (e.g. Japanese), git auto-merges them.
>>>
>>>  Sometimes there conflicts, but not all the time.
>>>
>>>  For example, when a new file is added to English, it just gets
>>>  auto-merged into Japanese. But all new changes must be
>>>  human-controlled, translated.
>>>
>>>  Is there a way to force git always generate a conflict, even if
>>>  changes could be auto-merged?
>>
>>  I am not sure what the workflow here and if it makes sense. When
>>  you have a file, "chapter47.txt", whose original is in English, the
>>  translation projects (there are N of them, one for each language)
>>  will have their own "chapter47.txt" that has translated text in the
>>  same place? It looks to me that, working that way, the project for
>>  translating into e.g. Japanese have no way to keep any of the
>>  original English version, in which case why are you even "merging"
>>  the English version in the first place?
>>
>>  I would have understood if the original "chapter47.txt" is translated
>>  into "chapter47_ja.txt" and friends, like "chapter47_fr.txt", all of
>>  which sit next to the original "chapter47.txt". Then merging an
>>  updated version of the original from time to time would make perfect
>>  sense to me---that would give you a way to see what changed in the
>>  original (e.g. "git show --first-parent -m -p master chapter47.txt")
>>  to guide you find the places you would need to make corresponding
>>  changes to the variant of your language, e.g. "chapter47_??.txt".