Experimental database dumps changelog

[pietroalbini]

This is a low-traffic issue tracking all the changes happening to the experimental database dumps. We recommend subscribing to this issue to get notified whenever we make some changes to the contents of the dumps.

[pietroalbini]

The next crates.io deploy (happening in the next few days) will include the following changes to the database dumps:

• PR Drop crates.textsearchable_index_col from the export #3612: The textsearchable_index_col column will be removed from crates.csv, as that column is an implementation detail of crates.io's search. Users importing the database dumps into a PostgreSQL database will not be affected by this change, as a trigger will populate that column at import time.
• PR Include only the last 90 days of downloads in our database dumps #3611: The version_downloads.csv file will only include the last 90 days of data instead of full day-to-day historical data. Cumulative download counts are still available in crates.csv and versions.csv.
• PR Remove version_authors and clean up old data #3549: The version_authors.csv file will be removed, as that data was deleted from the crates.io database too.

We also plan to make the following changes in the future:

• Issue Archive old entries in version_downloads table #3479: all the data from version_downloads.csv will be moved out of the database dump into separate files, one for each day. This will allow clients interested in this data to download it separately.

[Turbo87]

Two relevant changes were just deployed:

• database: Add checksum field #5077 and
• Stop saving badges in the database #5074

[Turbo87]

• database: Drop badges table #8155 will delete the badges table

[Turbo87]

• Extract crate_downloads table #8232 added a new crate_downloads table, which is supposed to replace the crates.downloads column soon. this was done for performance reasons to reduce the amount of bloat in the crates table from the regular downloads column updates. at the moment the data should be in sync, but if everything works out we will stop writing to the crates.downloads column in the near future and eventually remove it.

[Turbo87]

• as mentioned in the last update, worker/jobs/downloads/update_metadata: Remove crates.downloads column writes #8295 is going to disable writes to the crates.downloads column. we will keep the column around for now to avoid unnecessary schema churn, but once the system has shown the expected performance benefits we will most likely remove the column completely.

[Turbo87]

• once Remove crates.downloads column #8233 is merged and deployed it will remove the crates.downloads column. please us the crate_downloads table instead.

[Turbo87]

• Add default_versions database table #8484 will introduce a new experimental default_versions table with a mapping from crates to their "default" version, that will be shown by the frontend and used in e.g. reverse dependency queries.

[Turbo87]

• Create db-dump.zip file too #8748 added an experimental ZIP file artifact at https://static.crates.io/db-dump.zip. this file has the advantage of not having to decompress the entire file if you only need access to a certain database table CSV file. compared to the tarball the ZIP file does not have a top-level datetime path prefix, otherwise the files should contain the exact same data.

[Turbo87]

• Fix build metadata race condition #9756 added a num_no_build column to the versions table
• Exclude num_no_build from versions.csv db-dump #9786 changed it to be "private" since it is reasonably easy to reconstruct it locally if needed (thanks @dtolnay)

in other words: no changes (except for the few days in between the PRs) and sorry for forgetting to mention the change here 🙈