ENH: add PostgreSQL and MySQL support for pandas.io.sql.get_schema

[adamgreenhall]

Simple changes to add PostgreSQL syntax to pandas.io.sql.get_schema. Added data type cases for get_sqltype() and a columns "flavor" for get_schema().

The logic of get_schema seems simple enough at the moment (though it still doesn't support Oracle). Should it be replaced by SQLAlchemy integration?

ref: #4163
old pr: #4574

[jreback]

@adamgreenhall

yeh....have an open issue to disable that test, kind of flaky; ignore for now

[hayd]

I'm going to cherry-pick this when updating the branch.

[hayd]

I should have suggested before adding some tests for this. (At the moment the legacy tests are quite flaky and it's pretty WET and messy, so this may actually be non-trivial). Saying that, I think it's worth adding this as is too, thanks for putting it together.

One of the things coming soon (ish) is making it as simple as subclassing the tests with postgres engine/connection, which will be soo much better.

[jreback]

@hayd u r incorporating this?

[hayd]

@jreback yes, but I'm not sure how going to test all this legacy stuff yet (support for different engines is easier), atm no tests for this legacy (aka not sqlalchemy) postgres... I've just left what was there before.

Perhaps solution is to test using the connection against the engine version, not sure how that'll work yet.

[jreback]

@hayd to be honest, if you new stuff is good to go and covers sufficient test cases I would just go with that, maybe even move legacy engines temporarily for 0.13 so they at least 'work' and just migrate to new

[ghost]

@hayd, @adamgreenhall , superceded by #5950?

[jreback]

@hayd @jorisvandenbossche is this superceded? close?

[jorisvandenbossche]

I am not really familiar with this PR, but:

• we decided to not support postgresql in legacy mode, so only with sqlalchemy
• get_schema is deprecated

so judging from only the title of this PR (in the diff seems a lot more) this should be closed.

@adamgreenhall I am going to close, but if you think there is functionality in this PR that is not available in the new sqlalchemy stuff, we can reopen it.