NEWBY question - Pushing strings to Weblate

[JeroenVeldmeijer]

Hi everyone,

Please forgive me if the answer is very clear and to the point yet I'm 100th newby asking the same question.... if so, please help me find the answer by sharing the details / page / discussion. I couldn't find it easily but that could also be my newness to Github too. :(

We’re currently setting up translations for our e-commerce website (based on Sylius) using Weblate, and I’m running into some confusion — particularly around how the number of strings is determined and how to correctly offer content to Weblate via the API.

This is important for us because it impacts our decision between self-hosting and choosing a cloud plan (which is partly based on the number of strings). But there are several things I’m not clear on:

What exactly counts as a string?
Do strings refer to:

• Sentences?
• Paragraphs?
• Text fragments (e.g. headings, labels)?
• Or just the rows in de DB (ranging from words, sentences, full text (1000+ words)?

Do translations count toward the total number of strings, or is the count based solely on the source/original language?

The other thing I'm not clear on is the degree we should take control over how strings / content is offered to weblate. So, when using the API to push our original content from Sylius to Weblate, are we supposed to:

• Send full texts, or
• Break them into smaller parts that represent individual strings?

In other words, is it our responsibility to determine and manage the “string granularity” before sending content to Weblate?

We’d love to hear from anyone who has set up Weblate in a similar context and ran into the same issues and found a way to work around it. Any advice or examples of how you structured your integration or counted your strings would be hugely appreciated!!

Thanks in advance!
Jeroen

[emmapeel2]

What you do is you setup a repository with your translations, and make weblate follow it with a hook. So when you make changes to your repository, weblate will receive a message and update.

In my experience, I only send changes to weblate for the English strings, and let Weblate update the different translations. This way I avoid conflicts between the repository and weblate when they both try to edit the same file.

About the strings: One string is an arbitrary translation unit created by your translation framework. In my experience, the closest they are to 'one string == one sentence' the better, as very long strings will make it difficult to reuse the translations and are more difficult to update when there is a small change. We have some components that generate a very long string with several sentences and is horrible to translate/update/review.

[JeroenVeldmeijer]

Thanks @emmapeel2 ! Breaking up the content to sentences makes it rather complicated considering it needs to be reconstructed back just as the original content would be built up regarding formatting (headings, bullet points, etc.).

Again, not that familiair with Weblate so excuse me beforehand for silly questions - considering the described challange wouldn't it be most efficient if we would deactive manual editing of the content in the ecommerce system and add the content in Weblate incl. formatting and push it from Weblate to the ecommerce system only. Changes in the original content in Weblate activates 'checking procedures' (eg. content / sentence X is changed in language 1 and needs to be proof read / checked again / edited in language 2, 3, etc.) and after this the checked content (original content is already right after the changes) is pushed to the ecommerce system.

[emmapeel2]

But your platform should have a way to gather all those manual changes and generate a translation file, in one of several formats.

You first need to implrment internationalization on your platform, and then you feed the translation file generated by your system to weblate.

[JeroenVeldmeijer]

Thanks @emmapeel2 ! I checked and it uses YAML files to store translations. So we should work with that, thanks. I hope I'm getting closer !