[Cloud Spanner] Fix order of index columns

[naoina]

On Cloud Spanner, columns order of an index should be the same as the definition because it is important.

Probably you can reproduce it on Linux as follows using cloud-spanner-emulator and spanner-cli.

docker run --rm --net=host -d gcr.io/cloud-spanner-emulator/emulator:latest
sleep 5
curl -s localhost:9020/v1/projects/dummy/instances --data '{"instanceId":"dummy"}'
cat <<EOF | SPANNER_EMULATOR_HOST=localhost:9010 spanner-cli -p dummy -i dummy -d dummy -t
CREATE DATABASE dummy;
CREATE TABLE test (
    id INT64,
    key_a INT64,
    key_b INT64,
    key_c INT64,
) PRIMARY KEY (id);
CREATE UNIQUE INDEX test_index ON test(id, key_a, key_b, key_c);

SELECT
  c.INDEX_NAME, c.INDEX_TYPE, ARRAY_TO_STRING(ARRAY(
   SELECT COLUMN_NAME
   FROM INFORMATION_SCHEMA.INDEX_COLUMNS
   WHERE TABLE_NAME = c.TABLE_NAME AND INDEX_NAME = c.INDEX_NAME AND INDEX_TYPE = c.INDEX_TYPE AND COLUMN_ORDERING IS NOT NULL
   ORDER BY COLUMN_ORDERING ASC
 ), ", ") AS columns,
 ARRAY_TO_STRING(ARRAY(
   SELECT COLUMN_NAME
   FROM INFORMATION_SCHEMA.INDEX_COLUMNS
   WHERE TABLE_NAME = c.TABLE_NAME AND INDEX_NAME = c.INDEX_NAME AND INDEX_TYPE = c.INDEX_TYPE AND COLUMN_ORDERING IS NULL
   ORDER BY INDEX_NAME ASC
 ), ", ") AS storing_columns,
  i.PARENT_TABLE_NAME, i.IS_UNIQUE, i.IS_NULL_FILTERED, i.INDEX_STATE
FROM
  INFORMATION_SCHEMA.INDEX_COLUMNS AS c
INNER JOIN INFORMATION_SCHEMA.INDEXES AS i ON i.TABLE_NAME = c.TABLE_NAME AND i.INDEX_NAME = c.INDEX_NAME
WHERE
  c.TABLE_CATALOG = '' AND c.TABLE_SCHEMA = '' AND c.TABLE_NAME = "test"
GROUP BY c.TABLE_CATALOG, c.TABLE_SCHEMA, c.TABLE_NAME, c.INDEX_NAME, c.INDEX_TYPE, i.PARENT_TABLE_NAME, i.IS_UNIQUE, i.IS_NULL_FILTERED, i.INDEX_STATE;
EOF

18f0c531b197fca6f7a41e9756a50f25d7dbdcd169d52affdea360a4c4844756
{"name":"projects/dummy/instances/dummy/operations/_auto0","metadata":{"@type":"type.googleapis.com/google.spanner.admin.instance.v1.CreateInstanceMetadata","instance":{"name":"projects/dummy/instances/dummy","state":"READY"},"startTime":"2021-02-17T11:32:47.046560Z","endTime":"2021-02-17T11:32:47.046563Z"},"done":true,"response":{"@type":"type.googleapis.com/google.spanner.admin.instance.v1.Instance","name":"projects/dummy/instances/dummy","state":"READY"}}+-------------+-------------+-------------------------+-----------------+-------------------+-----------+------------------+-------------+
| INDEX_NAME  | INDEX_TYPE  | columns                 | storing_columns | PARENT_TABLE_NAME | IS_UNIQUE | IS_NULL_FILTERED | INDEX_STATE |
+-------------+-------------+-------------------------+-----------------+-------------------+-----------+------------------+-------------+
| PRIMARY_KEY | PRIMARY_KEY | id                      |                 |                   | true      | false            | NULL        |
| test_index  | INDEX       | key_c, key_b, key_a, id |                 |                   | true      | false            | READ_WRITE  |
+-------------+-------------+-------------------------+-----------------+-------------------+-----------+------------------+-------------+

[codecov]

Codecov Report

Merging #298 (910568c) into master (c3cffef) will decrease coverage by 0.07%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #298      +/-   ##
==========================================
- Coverage   66.54%   66.47%   -0.08%     
==========================================
  Files          25       25              
  Lines        4101     4101              
==========================================
- Hits         2729     2726       -3     
- Misses       1089     1091       +2     
- Partials      283      284       +1     

Impacted Files	Coverage Δ
datasource/datasource.go	40.14% <0.00%> (-2.12%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c3cffef...910568c. Read the comment docs.

[k1LoW]

Hi @naoina! Thank you for your commit!!

Oops. This is a bug...

https://cloud.google.com/spanner/docs/information-schema#information_schemaindex_columns

Column name	Type	Description
👍 ORDINAL_POSITION	INT64	The ordinal position of the column in the index (or primary key), starting with a value of 1. This value is NULL for non-key columns (for example, columns specified in the STORING clause of an index).
👎 COLUMN_ORDERING	STRING	The ordering of the column. The value is ASC or DESC for key columns, and NULL for non-key columns (for example, columns specified in the STORING clause of an index).