sebastianosinski / grdbsnapshottesting Goto Github PK

hello @SebastianOsinski,

I just wanted to tell that https://github.com/groue/GRDBSnapshotTesting is now an "official" member of the GRDB family, expected to evolve and be maintained in the long run along with GRDB.

I also wanted to thank you for your excellent idea of providing convenience methods for copying databases from test fixtures. They are now provided by GRDB itself: DatabaseQueue.inMemoryCopy(fromPath:configuration:).

That's it :-)

Hello @SebastianOsinski,

Thanks for GRDBSnapshotTesting, it works like a charm 👍 I wanted to suggest a change (two, actually).

Database snapshots currently contain internal SQLite and GRDB objects:

======== TABLES ========
CREATE TABLE grdb_migrations (
  identifier TEXT NOT NULL PRIMARY KEY
)

CREATE TABLE sqlite_sequence(
  name
  seq
)

======== DATA ========

## grdb_migrations
...

## sqlite_sequence
...

I'm wondering if those objects are useful for the tests.

As internal objects, they can change, without notice, from one version of SQLite or GRDB to another.

I'm also worrying that those objects make it difficult to generate snapshots without relying of the initial snapshot generation (*).

If you agree that we could do without them, then why not exclude them from the snapshot? It is easy to do so: all SQLite Internal Schema Objects have the sqlite_ prefix. All GRDB internal objects have the grdb_ prefix (not publicly documented, but I know it firsthand ;-)).

(*) On that second question, I'm also wondering if sorting objects by name wouldn't be a net improvement. Sorting objects by name would have another benefit: by easing the comparison between two subsequent schemas with a simple diff, it helps deciding if a snapshot is correct or not.