Launchpad Translations Reporting API
This Launchpad enhancement proposal discusses extending the functionality of Launchpad Translations by developing an API which will expose the necessary interfaces to allow read access to translations-related data from Launchpad through launchpadlib for reporting purposes.
As an Ubuntu Translations Coordinator
I want to be able to report on the status of Ubuntu translations at the distro level (only default installation source packages)
so that I can present the data to the community in a way they can see the real status of translation, track their work, set translation goals, and in short, improve the translation coverage in all languages
As an OEM Engineer
I want to be able to create an automated report on the translation coverage of a set of source packages in a given set of languages
so that I can inform a customer and assess whether additional translation work is required before shipping
As an Ubuntu translator
I want to be able to obtain translations data in an automated way
so that I can present and use the data in alternative ways that will help my team better organize translation work
Due to other higher priority development work, the Translations component has been lacking a full API for quite some time now.
The need of an API arises now from the necessity of better reporting the status of Ubuntu translations. The main benefit accessing data through launchpadlib can provide to Ubuntu is on the statistics reporting side: currently the main access point to translating Ubuntu is at http://translations.launchpad.net/ubuntu. This view offers statistics to all translatable packages in the main and restricted repositories, regardless of distribution. While complete, it does not provide information on how well translated Ubuntu is as a distribution. This information can be very useful to Ubuntu translators in assessing what needs to be translated and on which packages they should concentrate their work. This can also be used to track progress, set up goals, and in short, as an aid to building community around translations. In the same way, this can be useful for commercial customers to assess translation coverage when evaluating to ship Ubuntu in a given language.
As the current focus of Launchpad development is on upstream integration, and as there is no planned UI work in that area, an API appears to be the most adequate way to allow accessing the Launchpad Translations data and present it in alternative ways to suit the stakeholders' needs.
(David: as the drafter of this proposal, I'm regularly in touch with all of the stakeholders)
The new behaviour must provide a way to access translations data from Launchpad, either in read (or depending on the value) or write mode through launchpadlib, without the need for an admin to query the database directly.
It must not provide any additional functionality or a way to circumvent any security privileges built in the layers above (e.g. Launchpad UI)
This section will be used to list every separate aspect of reporting, with DATA lists as the proposals for the data to be exposed.
Retrieving the list of all languages available in Launchpad. For each language the following data would be required:
- Language code
- Language name in English
- Language state (visible or not)
- Number of contributors
Plural forms [optional, for checking language support]
Text direction [optional]
The rest of the data at https://translations.launchpad.net/+languages/$LANG [optional]
It would be also desirable to have the ability of querying a single language as well.
Retrieving the list of all templates available for a series in Launchpad. For each template the following data would be required:
- Template name in Launchpad
- Date last updated
- Domain name
- Iscurrent (active or not)
- Source package associated
- Source package repository (option - to avoid a new query on source package)
- Is it used in language packs
- The list of POFiles associated with this POTemplate
In which languages there is a translation available [optional]
It would also be nice to be able to query a template from a series, based on the Launchpad template name.
Retrieving the list of all PO files for a template in a series.
- Language code
- Number of translated messages
- Number of unreviewed messages
- Number of changed messages
- When the translation was last updated
- Who the last translator was
We already have a few exported attributes for source package, but to help in identifying translations moved to univers we also need to know the archive/repository for this sourcepackage (https://api.launchpad.net/+apidoc/#source_package)
- Archive/repository name
Danilo's thoughts on structuring the work for the reporting part:
Exposing API for single PO file statistics - a shared interface IRosettaStats will need to be exported through API, so all the other objects which are based on it will get the same API as well.
Devising a good API for a single PO template
- Other than the simple properties, the most interesting bit here is how to get statistics for a bunch PO files attached to a template. If you don't do anything but just export a single method to list out all the PO files, you'll be doing one API call for each of the languages for a PO template and that's going to be very slow. Eg. to get full status export for Ubuntu out of LP you'd need to do some ~400k API calls.
A much better solution to that is to provide a method which gives you all related PO file statistics: e.g. getPOFilesWithStatistics() which returns tuples of form (POFile, translated_count, changed_count, unreviewed_count, untranslated_count) directly. FWIW, I'd keep this a separate step as well.
IPOTemplateS(ubs)et API - as the final step, I'd do an export of IPOTemplateS(ubs)et APIs for all the different objects that have them. As another part of this step, you'd probably want to do the similar to above and provide a getPOFilesWithStatistics(language) method on them.
The idea behind this workflow is simply to use a gradual approach, which will be easier to get right and get landed, without ending up with a too large a branch to review or follow.
Although the intention is to propose a full API, the two main success criteria are the implementation of API functions for reporting and the imports queue, so we will know that we are done when we can report on the status of Ubuntu translations at a distro level (i.e. on a given set of packages, for all Launchpad languages) by having sourced the data trough launchpadlib.
The success can be measured in how seamlessly this data can be obtained and in whether all requirements in terms of exposed data have been implemented.