fix: types reflection - #33
AVaksman merged 16 commits into
Conversation
Sorry, something went wrong.
| "FLOAT64": types.Float, | ||
| "INT64": types.BIGINT, | ||
| "NUMERIC": types.DECIMAL, | ||
| "NUMERIC": types.NUMERIC(precision=38, scale=9), |
There was a problem hiding this comment.
Constant precision and scale
Sorry, something went wrong.
| size = int(col[1][7:end]) | ||
| type_ = _type_map["STRING"](length=size) | ||
| else: | ||
| type_ = _type_map[col[1]] |
There was a problem hiding this comment.
While reflecting column types, we're reading strings like STRING(size) from information_schema. Such a string can be parsed to get the column size - to set it into SQLAlchemy reflected type.
@larkee, I assume, something similar is required for BYTES (you mentioned their MAX size in the dialect), so I'm looking forward for BYTES types test.
Sorry, something went wrong.
|
@larkee, as this PR is based on #32 and 32 is based on #24, the branch includes all the changes of my last three PRs. With all of this I have only three reflection tests failing. Most of the skipped tests are related to the named schemas feature, so they are going to stay skipped. Thus, we're almost there with reflection I suppose.
|
Sorry, something went wrong.
|
@larkee, here I'm also adding results testing group into the test suite, as all the tests of the group seem to be processed: PercentSchemaNamesTest are test cases to test if % symbol can be used in column, table, schema names. % can't be used in Spanner identifier, so leaving the tests skipped. ServerSideCursorsTest - I didn't find any mentions of server cursors in Spanner, so I assume it's not supported, and the test case can stay skipped.
|
Sorry, something went wrong.
This PR is based on #32.