Commit graph

4 commits

Author SHA1 Message Date
Pat Okeeffe
5c74b64722
A bug fix roundup (#2242)
* fb-1940 re-implemented some changes that got missed private-public

* fb-1939 fixes to between + decimals

* fb-1935 - avg() on and id type + fixed some tests

* fb-1953 add min/max for string types

* fb-1938 - remove internal_type column from show columns

* fb-1964 - fix space_used in fb_cluster_nodes to be int

* fb-1996 - make sure all Idents that are being used as object references to schema objects are lowercased

* fixed failing test

* added some missed changes

* fb-1969 found another case issue with identifier used for column idents
2023-02-13 12:28:34 -06:00
pokeeffe-molecula
c304991e9f implemented extract ddl; tightened up type related stuff (#2329)
* implemented extract ddl; tightened up type related stuff

* added some test coverage

* review feedback

(cherry picked from commit f62313762c)
2022-12-12 09:01:20 -08:00
Travis Turner
444d4804ec Fix formatting in CLI results with custom SQLResonse.UnmarshalJSON (#2305)
* Fix formatting in CLI results with custom SQLResonse.UnmarshalJSON

When I started this, it was meant to be a quick fix to address the confusing
result formats we were seeing in the CLI. For example, all large integer values
were displayed in scientifc notation. This is because we were passing the result
types from JSON (in this case, float64) into pretty print. Similarly, `IDSets`
and `StringSets` where being printed using the default go Stringer for the types
[]int64 and []string respectively.

I started by writing a customer UnmarshalJSON() method for the `SQLResponse`
type. Part of this (the part which converts data types based on header types)
was already being used in dax tests, so this just formalizes that logic as part
of the `SQLResponse` type.

Then I realized that the sql3 tests (run against the `sql3` package) were
failing because sql3 is not actually returning the `IDSets` and `StringSets`
types. A future task is to formalize return types, define them, and modify sql3
to return them. Once that is done, we can remove the "typed" switch in the
`SQLResponse` json unmarshaller.

Another significant change is the modification to the `ExprDataType` interface:
```
type ExprDataType interface {
	exprDataType()
	TypeName() string
	TypeDescription() string
	TypeInfo() map[string]interface{}
}
```
I added two more methods in order to distinguish between a type (`DECIMAL`), its
description (`DECIMAL(2)`), and its type info (`"scale": int64(2)`). Currently,
the description can be used as the field definition in a CREATE TABLE statement,
but we may want to re-think that. Also, Decimal is the only type currently using
TypeInfo.

Finally, I tried to consilidate things around `dax.FieldType` instead of
comparing against parser types outside of sql3. We still have some sql3 parser
and planner types lurking about, but we can address those in future commits.

* Add some test coverage

* smoke test expected INT, now int

* minor fixes

* Introduce WireQueryResponse and related types

This also changes dax.FieldType to dax.BaseType.

* Populate WireQueryResponse correctly

Currently this is in the http handler, and in the queryer.

* Convert sql3 and dax tests to expect pilosa.WireQueryField in results

* fix PQL tests in the SQL defs

* Address a few of the skipped sql tests in dax

(cherry picked from commit f4385df2cf)
2022-12-12 09:01:20 -08:00
Travis Turner
53170cef64 Move SQL3 test definitions into sql3/test/defs package (#2289)
This so other packages can import and use them in their own tests.

(cherry picked from commit eec278893a)
2022-11-15 11:33:35 -08:00
Renamed from sql3/sql_defs_between_test.go (Browse further)