Conversation
rfk
left a comment
rfk
left a comment
There was a problem hiding this comment.
add **kw to capture other unused args.
Hrm, I'm not sure about this, it seems like it might risk just masking future errors rather than failing loudly with a signature mismatch, and having them surface only later as incorrect data written into the db. @jrconlin thoughts?
Existins users of this codebase will need to run a db migration to get the new fields, will the existing migration from the tokenserver codebase work correctly here?
I view the StaticNode to be kind of a special case. I'll note that we had already added the
I see that the tokenserver migrations include both of those fields, so I presume that if the migration process works correctly, the table would have already been changed. |
|
sigh. looks like the migration script does not fire automatically. I'll see if I can work out what's going on with alembic or at least add instructions to existing syncstorage users. |
* debugging code * Fix get_collection_id and add some debugging cruft * Fix a few more things * Fix a few more errors * Fix some more stuff * Example of how to convert a timestamp properly and some more bugfixes * Fix another error * Fix another timestamp * wip fill in transaction ids * checkpoint * fix nanoseconds/cleanup * sql_request * some error handling cleanup (converting google_spanner1::Error to DbError) * wip * wip * Don't shadow q; don't add a , if we're adding an empty string; manage lifetimes properly in the comma function * Fix put_bso * kill * Parse timestamps as rfc3339, not i64 * Fix get_bso * Fix get_bso_timestamp * Switch all timestamp formatting to to_rfc3339 function, fixing bso_successfully_updates_single_values * Return a map of collection names to counts, not collection ids to counts * Fix spellling error * Fix delete_collection * Fix delete_bsos_in_correct_collection * ORDER BY not SORT BY * Fix get_collection_timestamps * Fix get_collection_usage * Fix create_collection when the database is empty * use row_count_exact * phrasing * Fix get_bsos_sort * Make all of the tests pass * fix up transaction handling: - delete_storage lacks a collection so there's no lock_for_xxx called for it - create_collection may be called within lock_for_write (before the transaction is created) so, like the python version, if no txn is present and one is asked for, we implicitly create a r/w txn * Actually fix conflicts * Switch to eprintln * followups
4 participants
**kwto capture other unused args.Closes #191