sillsdev / overcrowdin Goto Github PK

In a large project, it can be tedious to keep track of which localizable files have been added or deleted. It would be nice to have a one-step process to add, update, and delete files in Crowdin.
Basic design:

• Poll Crowdin for a list of existing files (https://support.crowdin.com/api/info/)
• Build a list of local files specified in crowdin.json
• Combine these two lists to create three lists: add (incl. new folders), update, delete
• Perform these three operations (perhaps we should delete only when the --prune option is specified)

See https://support.crowdin.com/configuration-file/#usage-of-wildcards
Support:

#........your project configuration........
"files": [
  {
    "source" : "/**/*.*", #upload all files that  the base_path contains
    "translation" : "/**/%two_letters_code%_%original_file_name%",
    "ignore" : [
      "/**/%two_letters_code%_%original_file_name%", #ignore the translated files
      "/**/*Tests.*"                                 #ignore test files
    ]
  }
]

If crowdin.json contains any unsupported options, warn the user with a message like Unsupported option: X. This should not affect the errorlevel, but the warning should be displayed regardless of whether -verbose is specified.

Calls to ExportTranslation time out for large projects. The fieldworks project has 7500+ strings in 22
languages, which take ~22 minutes to build, but the export command times out after ~100 seconds. This may need to be adjusted in https://github.com/crowdin/crowdin-dotnet-client Crowdin.Api.Client.ExportTranslation's web request.

Implement the options documented at https://support.crowdin.com/configuration-file/#additional-options-for-xml-files (additional details at https://support.crowdin.com/api/add-file/). Options include translate_content, translate_content, content_segmentation, and translatable_elements.

Crowdin does not accept uploads of more than 20 files at once. To support larger uploads, we must batch them and possibly pause between batches.
Edit: We should upload files one at a time so that an entire Update operation doesn't fail because a single file doesn't exist (or an entire Add operation fail because a file does exist).

When adding or updating files, filter and remove resx files that have no localizable strings. Localizable strings are the content of the <value> elements of <data> elements whose name attribute either contains no dots (as in a project's Strings.resx) or whose name attribute ends in .Text, .AccessibleName, or .AccessibleDescription.

I've seen at least one instance of Crowdin prompting to localize a binary resource (base 64 byte array), but can't seem to find it again to determine the context. .Icon values are hidden in Crowdin.

API v1 is no longer supported as of 2022.

Crowdin will not allow uploading files to folders that don't exist. We need to

• add the Add Directory API method (even if we don't expose it on our CLI)
• add directories before adding files to them

When adding or updating files, filter and remove xml files that have no localizable strings. For minimum viable product (MVP), find any matches for the XPaths specified in translatable_elements, if any. Bonus points for checking that the specified attributes and elements have content.

Implement calls to https://support.crowdin.com/api/pseudo-download/ and https://support.crowdin.com/api/pseudo-export/

Currently, the only wildcard characters we recognise are ** at the end of a path (search directories recursively) and single *'s in filenames. We would like to be able to match any number of ** and * anywhere in the path. Specifically, we would like to "ignore" : [ "/Src/**/*Tests/**/*.resx" ].

Files specified using the command line option -f, -file can have paths that are absolute or relative, or that contain arbitrary directory separator characters. These paths are presently used as keys without any modification or vetting. It would be nice if they were somehow normalized. Some options:

• remove any Windows drive letters (e.g. C:) and normalize to Unix directory separators
• use only the filename as the key (this may require less code to enforce, but also gives users less flexibility)