Load() calculate hash of every table at the end and I don't know why

[slsyy]

v3.11.0

The last function call for Load():

testfixtures/testfixtures.go

Line 460 in b1fe75b

return l.helper.afterLoad(l.db)

and postgres implementation

testfixtures/postgresql.go

Lines 362 to 376 in b1fe75b

	func (h *postgreSQL) afterLoad(q queryable) error {
	if h.tablesChecksum != nil {
	return nil
	}
	h.tablesChecksum = make(map[string]string, len(h.tables))
	for _, t := range h.tables {
	checksum, err := h.getChecksum(q, t)
	if err != nil {
	return err
	}
	h.tablesChecksum[t] = checksum
	}
	return nil
	}

Basically it iterates over the whole database (in my case it is 100+ tables) and calculate md5 hash of each table for no reason. Do I miss something or maybe it can be removed? In my case it slows down tests in noticeable way

[slsyy]

Oh I see, it make sense for subsequent Load calls.

Do you think it make sense to make some flag to disable this behavior? In my case I use Load only once per loader, because I use https://github.com/DATA-DOG/go-txdb , which perform cleanup automatically

[andreynering]

Hi @slsyy,

We already have DangerousSkipCleanupFixtureTables. Can you try and see if it works for you?

I'm not sure if it also skips the checksum computing for all tables, but if not it would make sense to fix that IMO.

[slsyy]

@andreynering this flag control that piece of code:

testfixtures/testfixtures.go

Lines 420 to 430 in b1fe75b

	if !l.skipCleanup {
	for _, file := range l.fixturesFiles {
	modified := modifiedTables[file.fileNameWithoutExtension()]
	if !modified {
	continue
	}
	if err := file.delete(tx, l.helper); err != nil {
	return err
	}
	}
	}

where I want to skip this:

testfixtures/testfixtures.go

Line 460 in b1fe75b

return l.helper.afterLoad(l.db)

Reason: In my code I use this library in such a way:

	fixtures, err := testfixtures.New(...)
	handleErr(t, err)
	handleErr(fixtures.Load());
	// and fixtures object dies here

so I basically don't need all of those hashes, because they won't be used anyway, if I don't call Load() more than once. With txdb anyway it doesn't make sense as tx.Rollback revert all the changes

I can raise a PR, if it make sense

[andreynering]

@slsyy I'm not so sure if we could do the fix for the existing DangerousSkipCleanupFixtureTables flag, or if we'd need a new flag this. I'd need to revisit the code to be sure.

Anyway, feel free to open a PR with a proposal if you want to!