mirror of
https://github.com/soxoj/maigret.git
synced 2026-09-22 17:54:55 +02:00
SoxojandGitHub
ec72abb2d2
Moves 395 sites from the `Discourse` engine to `DiscourseJson`, which checks
`/u/{username}.json` instead of the HTML profile page.
The JSON answer carries the fields socid_extractor already parses — uid, trust
level, badges, registration and last-seen dates: sixteen fields against four,
in the same single request.
Measured before the move, on all 464 Discourse entries: 403 serve that JSON to
a guest, and a name that cannot exist answers 404 with
`{"errors":[…],"error_type":"not_found"}`. Each of the 395 was then run through
maigret on the converted entry — the claimed name reports CLAIMED, three
impossible names report AVAILABLE.
Eleven are more accurate than before, not merely richer: on
`community.brave.com`, `forums.sketchup.com`, `community.smartthings.com` and
eight others, an account that exists answers `/u/<name>.json` with its user id
while `/u/<name>/summary` answers 404 — so the HTML engine has been reporting
real accounts as available.
Left on the HTML engine: 9 entries that answer 200 with HTML for a real member
and for an impossible name alike, 9 that answer 403, 21 that did not answer at
all, and `discuss.ardupilot.org`, the one entry where the JSON route proved
less reliable than the HTML one.