First steps towards AR 5 support on Postgres

[md5]

Also includes an update to the version range to limit to AR < 6 (cf. SchemaPlus/schema_monkey#11) as well as some updates to the specs that relate to the order that AR 5 returns columns for polymorphic references.

[coveralls]

Coverage remained the same at 100.0% when pulling b89016a on md5:activerecord-5.0 into fa201bb on SchemaPlus:activerecord-5.0.

[coveralls]

Coverage remained the same at 100.0% when pulling 30b2f09 on md5:activerecord-5.0 into fa201bb on SchemaPlus:activerecord-5.0.

[md5]

Most of the failures for Postgres look like this:

Failure/Error: ActiveRecord::SchemaDumper.dump(ActiveRecord::Base.connection, stream)
     NoMethodError:
       undefined method `[]' for nil:NilClass
     # ./lib/schema_plus/core/active_record/schema_dumper.rb:78:in `block (2 levels) in table'
     # ./lib/schema_plus/core/active_record/schema_dumper.rb:69:in `map'

Those are caused by the inline t.index calls in create_table that I mentioned in #9

[coveralls]

Coverage decreased (-0.2%) to 99.752% when pulling 180875f on md5:activerecord-5.0 into fa201bb on SchemaPlus:activerecord-5.0.

[md5]

Looks like the t.index statement will actually need to be parsed and turned into a SchemaDump::Table::Index instance (or something similar that outputs a new t.index string). Not sure whether it makes sense to add it to env.table.statements or have something like env.table.indexes.

[ronen]

The prepare_column_options fix looks fine of course. And I'm happy to accept it (despite the stil-failing build, since we can't expect any one PR to fix all the problems). But your other commits are all getting mingled in; especially 180875f. For cleanliness (a little bit for the commit history but mostly for the discussions) maybe make separate branches & PRs for each?

Regarding t.indexes: There actually already is env.table.indexes, see https://github.com/SchemaPlus/schema_plus_core/blob/970bf76/lib/schema_plus/core/active_record/schema_dumper.rb#L88-L100. So instead of adding m[:statements] I'd add m[:indexes]. But maybe move this discussion to a separate issue/PR?

Thanks again!

[md5]

@ronen Yeah, I got carried away trying to get the tests passing. Do you mind 427ab39 being in this PR?

[md5]

Or 30b2f09 for that matter

[md5]

I've dropped 180875f from this PR

I'll open an issue to discuss indexes_in_create separately.

[ronen]

In some highly principled world I suppose 427ab39 and 30b2f09 would be separate PRs. But they don't really need any discussion so let's just pretend this PR is called "first steps towards AR 5 support" and merge 'em :)