TODO: Add content
The data for the family visualizer is contained in .json files within the json/ folder. This data can be automatically updated using the scraper.py script located in the scrapers/ folder.
The scraper gets data for a specified family, which should be passed as an argument (python3 [path]/scraper.py FAMILY).
By default, the script gets the history of all languages and pairs in the family, as well as some other data used by the visualizer, for languages in the specified family. This data will be outputted to the {lang}.json files (history of each language and its pairs) as well as the {family}_pairData.json and {family}_transducers.json, which contain more generic data about each family (contributors, state, location, stems). The script can, however, also be run in "shallow" mode (-s or --shallow) to only get data for the {family}_pairData.json and {family}_transducers.json files, to avoid long runtimes.
To count contributions properly, the script uses a .mailmap file, to avoid commits made by the same person but under different names/emails being counted separately (e.g. commits by "John Doe" and by "J. Doe" would be counted separately). To update this file, the script has an "update mailmap" mode (-u or --updatemailmap) that outputs the emails that aren't on the .mailmap file, along with some data to help identify the authors. They must then be added manually, as the script can't figure out whether this committer is already in the file, but with a different email. Also, you could technically choose any name for a committer, but it is better to choose their GitHub username. If they don't have a GitHub account, the name they committed under most frequently should be chosen. If you are running this option solely to output the unknown committers, the option can be run along with shallow mode, as the only committers that are logged are those that would go in {family}_transducers.json, since the contributions are only counted for this file and this file is already updated with shallow mode. Moreover, there is currently a list of authors available in apertium-packaging. This list unfortunately doesn't store all emails for an author, so it can't be used for this project. However, that list and .mailmap are similar and should probably eventually be synced somehow.
Because of an issue with the GitHub API not working well with file renames, the script clones the repo for each language and pair (into the scrapers/git-repos/ folder), if they're not already present and updates them if they are, and gets the data by running git log in it (which does output commits past renames), which is slow, but there is currently no other option to get the full history of a dictionary. To avoid being slower than necessary, the scraper ignores commits that are already in the files and only gets data for the ones that aren't. This scraper should be updated if either the GitHub API is updated to support history past file renames or if the stats-service is updated to support history (see this issue).
Notes:
- The scraper prints some info about what it's working on at the time. This can be avoided using the quiet option (
-qor--quiet) - Some commits are skipped because their stems can't be counted (because of a syntax error in the dictionary, in most cases). To print some info about these, use
-vor--verbose. - Occasionally, the script will return an exception saying that the stats-service is updating, either in the
monoDataorpairDatafunction. While it's updating, not all the necessary data is available. If this happens, please try again later (or run the script for another family, as the stats-service updates progressively). - The data about which language is in each family is in
scrapers/families.jsonand when Apertium starts supporting a new language, it should be added manually to the file. A language should only be added if it both has a monolingual module and is present on the "transducers" table in the family's wiki page. - The
scrapers/iso3to1.jsonallows the script to convert the language codes from iso639-3 standards to iso639-1 standards. This is useful for file renames, as most renames were to convert the file name from iso639-1 to iso639-3 standards. The script relies on all languages needed being there, including the ones that don't have an iso639-1 code, which then just convert to themselves. If a language is, for some reason, not present on the file it should be added. - If run on Windows,
gitcould give such errors:error: unable to create file tests/morphotactics/prefixes/PRN.*: No such file or directoryand thenfatal: unable to checkout working tree'. This causes no issues for the script and is due to this issue - The scripts contained within the
scrapers/svn-oldfolder are the old scrapers that got the data from the svn repo. They shouldn't be run anymore.