ENH: support index=True for SQL io.sql.get_schema

[jorisvandenbossche]

pd.io.sql.get_schema can be used to see the schema 'create table' statement. This function should support including the index (which is actually the default for to_sql)

In [31]: df = pd.DataFrame({'col':[1,2,3]}, index=pd.date_range('2012-01-01', periods=3))

In [33]: print pd.io.sql.get_schema(df, 'test_schema', con=engine_postgres)

CREATE TABLE test_schema (
    col BIGINT
)

In [34]: print pd.io.sql.get_schema(df.reset_index(), 'test_schema', con=engine_postgres)

CREATE TABLE test_schema (
    index TIMESTAMP WITHOUT TIME ZONE, 
    col BIGINT
)

[trbs]

I can create a PR for this if wanted.

Would we want the patch (i assume so) to be backwards compatible with index=False as the default for get_schema or have the PR conform to the default index=True value as is used in other places else. (meanly to_sql)

[fangchenli]

After adding the index option, the returned default "index" name has double quotes. Is this expected?

df = pd.DataFrame({
            "one": [1, 2, 3], "two": [1, 2, 3]
        }, index=list("abc"))

schema_with_index = sql.get_schema(df, 'test', index=True, con=self.conn)

CREATE TABLE test (
	"index" TEXT, 
	one BIGINT, 
	two BIGINT
)

df.index.name = 'new_index'
schema_with_index_rename = sql.get_schema(df, 'test', index=True, con=self.conn)

CREATE TABLE test (
	new_index TEXT, 
	one BIGINT, 
	two BIGINT
)