V2

[josephbuchma]

Quite massive backward-incompatible refactoring :)
All drivers and journey cli are also updated.
This PR addresses all concerns discussed in #2 (in one way or another)

Changes

• Removed all "async" stuff.
• Added type for encapsulating migrations functionality for greater flexibility and performance.
• Changed driver registration and initialization approach, fixed Driver must be stateless or not a singleton #5
  • Removed Initialize method from Driver interface
  • Removed FlenameExtension from Driver interface
• Added driver.Locker interface, which can be optionally implemented to enable locking during migraitons.
  • Implemented for mysql driver
• Migrate now receives context.Context, and therefore can be cancelled.
• Added option to attach pre/post hooks for migrations.
• Added methods for applying / rolling back specific version.

[gravis]

A lot of things to review :)
Please give me a few days

[josephbuchma]

@gravis tests on master branch are failing

❯❯❯ make test                                                                   ⏎ master ✱
docker-compose run --rm go-test
Starting migrate_crate_1
=== RUN   TestCreate
--- PASS: TestCreate (0.24s)
        migrate_test.go:31: Test driver: postgres://[email protected]:5432/template1?sslmode=disable
        migrate_test.go:31: Test driver: mysql://root@tcp(172.17.0.9:3306)/migratetest
        migrate_test.go:31: Test driver: sqlite3:///tmp/migrate.db
=== RUN   TestReset
--- PASS: TestReset (0.73s)
        migrate_test.go:72: Test driver: postgres://[email protected]:5432/template1?sslmode=disable
        migrate_test.go:72: Test driver: mysql://root@tcp(172.17.0.9:3306)/migratetest
        migrate_test.go:72: Test driver: sqlite3:///tmp/migrate.db
=== RUN   TestDown
--- FAIL: TestDown (0.17s)
        migrate_test.go:110: Test driver: postgres://[email protected]:5432/template1?sslmode=disable
        migrate_test.go:140: Expected version 0, got 20171227092036
=== RUN   TestUp
--- FAIL: TestUp (0.05s)
        migrate_test.go:147: Test driver: postgres://[email protected]:5432/template1?sslmode=disable
        migrate_test.go:165: Expected version 0, got 20171227092036
=== RUN   TestRedo
--- PASS: TestRedo (1.03s)
        migrate_test.go:189: Test driver: postgres://[email protected]:5432/template1?sslmode=disable
        migrate_test.go:189: Test driver: mysql://root@tcp(172.17.0.9:3306)/migratetest
        migrate_test.go:189: Test driver: sqlite3:///tmp/migrate.db
=== RUN   TestMigrate
--- FAIL: TestMigrate (0.18s)
        migrate_test.go:229: Test driver: postgres://[email protected]:5432/template1?sslmode=disable
        migrate_test.go:267: Versions in db: [20171227092036 20171227092035 20171227092034 20171227092033 20171227092032 20171227092031 20171227092030 20171227092029 20171227092028 20171227092027 20171226235039 20171226235038 20060102150405]
        migrate_test.go:268: Expected version 0, got 20171227092036
FAIL
exit status 1
FAIL    github.com/db-journey/migrate   2.399s
?       github.com/db-journey/migrate/direction [no test files]
?       github.com/db-journey/migrate/driver    [no test files]
=== RUN   TestParseFilenameSchema
--- PASS: TestParseFilenameSchema (0.00s)
=== RUN   TestFiles
--- PASS: TestFiles (0.00s)
=== RUN   TestDuplicateFiles
--- PASS: TestDuplicateFiles (0.00s)
PASS
ok      github.com/db-journey/migrate/file      0.003s
?       github.com/db-journey/migrate/pipe      [no test files]
Makefile:7: recipe for target 'test' failed
make: *** [test] Error 1

[josephbuchma]

@gravis I added optional pre/post hooks for migraitons.
Useful for automating more complex migrations, and it also can be used for progress logging on client side (CLI).

[josephbuchma]

@gravis just want to note that I think I addressed all concerns discussed in #2 (in one way or another)

[josephbuchma]

Happy New Year !

Waiting for your feedback @gravis :)

[josephbuchma]

ping

[gravis]

Hi there. I'm back to work today, sorry for the delay.
I'll review your PR asap.
Thanks

[gravis]

LGTM.
Handle name is certainly not blocking.
Let's continue with changes in drivers, and we'll be good to merge

[gravis]

Shouldn't we name it Handler, instead of Handle?

[josephbuchma]

I don't like Handler b/c it's common name/pattern which usually refers to interface or function.
We can go conservative way and name it Migrate . I was simply looking for something that plays well with migrate. prefix and isn't counter-intuitive (at least), and Handle is the best what I came up with.

[josephbuchma]

round 2 @gravis 😄
This time it's a bit more radical. I decided to fix #5 and it went a bit further.
All drivers are up to date in respective PRs

[josephbuchma]

@gravis now I think I'm happy with results. From my perspective it's ready for merge, and drivers can be polished with follow-up PRs.

[gravis]

Thanks for your tremendous work!