CSV formatter

[Bogidon]

Yet Another Try at adding a CSV formatter (related: #34, #41, #154).

I went for a barebones formatter, opting out of creating a convention for encoding multiple languages in one CSV file. The format is similar to Apple's .strings except that language information is encoded in the file name:

• strings-en.csv
• strings-ro.csv
• etc.

Here's some sample CSV, pulled out from the test fixtures:

Key,Value,Comment
key1,value1-english,comment key1
key2,value2-english,
key3,value3-english,
key4,value4-english,comment key4

I was a little confused about where exactly I should be adding tests. Just let me know what I should fix and I'll jump on it!

[sebastianludwig]

Thank you very much for your work! At a first glance it looks very good, I'll take a closer look on the weekend 👍

[Bogidon]

Great, thanks!

[scelis]

Interesting. Thank you for this! I am curious, though. I would imagine that people using CSV would want the file to contain multiple translations instead of just one. Is there anyone out there with a use-case for CSV files that only have a single translation?

Thus, how useful is this current implementation? And if we decide to add multiple languages later, would that break anyone who decided to start using this one?

[Bogidon]

@scelis, yeah I'm starting to lean towards multi-language CSVs too. I went for single language because it is I need in my particular use case, but I can easily see how multi-language would make a lot more sense. Just to make sure we consider all options, two other solutions could be

1. a separate multi-language CSV formatter
2. a flag that toggles between multi-language and single-language. (though this is probably undue complexity and probably an anitpattern)

I'll update the PR for whatever we decide on.

[scelis]

I don't think we need an extra flag or an extra formatter. What do you think about using the existing --lang flag? That flag should support a comma-separated to support multiple languages. That way the single formatter could be used to create a single language CSV or a multi-language one. The only update this PR would need would be to add a header row describing the position of each language in the file. For example:

key,en,es,comment
hello,Hello,Hola,A greeting

[sebastianludwig]

Short and concise 👍 I only added two comments/questions.

[sebastianludwig]

If it's important to not have empty lines in the CSV, we should add a unit test for this

[sebastianludwig]

doesn't this also add the header as translation?

[sebastianludwig]

In general I like the idea of having a CSV formatter a lot. It's been requested multiple times. However I also think it should be a multi-language formatter. By default it should output all languages if the --lang parameter is omitted.

@Bogidon do you want to give the implementation a go?

(Sorry again for taking ages to react :-/)

[Bogidon]

Also sorry for leaving this up in the air for so long (my I18N project at worked got temporarily shelved). I'll definitely have a go on it, hopefully in the next few days.

[scelis]

Closing due to inactivity. Please reopen if you ever have time to revisit!