perf(db): Extend index on cards_properties to cover name and value

[ChristophWurst]

• Resolves: #

Summary

Search in in addressbooks dispatches a query like

SELECT DISTINCT `cp`.`cardid`
FROM `oc_cards_properties` `cp`
WHERE (`cp`.`addressbookid` IN (18))
  AND (`cp`.`name` IN ('FN'))
  AND (`cp`.`value` COLLATE utf8mb4_general_ci LIKE '%john%')
LIMIT 20

MariaDB picks card_name_index, which only covers name, and still has to scan 712 index entries and then do a table access.

{
  "query_optimization": {
    "r_total_time_ms": 0.144922892
  },
  "query_block": {
    "select_id": 1,
    "cost": 0.6448338,
    "r_loops": 1,
    "r_total_time_ms": 1.652976417,
    "temporary_table": {
      "nested_loop": [
        {
          "table": {
            "table_name": "cp",
            "access_type": "range",
            "possible_keys": ["card_name_index"],
            "key": "card_name_index",
            "key_length": "259",
            "used_key_parts": ["name"],
            "loops": 1,
            "r_loops": 1,
            "rows": 712,
            "r_index_rows": 712,
            "r_rows": 712,
            "cost": 0.6448338,
            "r_table_time_ms": 1.404896605,
            "r_other_time_ms": 0.240330463,
            "r_engine_stats": {
              "pages_accessed": 1426
            },
            "filtered": 100,
            "r_total_filtered": 0.140449438,
            "index_condition": "cp.`name` = 'FN'",
            "r_icp_filtered": 100,
            "attached_condition": "cp.addressbookid = 18 and cp.`value` collate utf8mb4_general_ci like '%user0%'",
            "r_filtered": 0.140449438
          }
        }
      ]
    }
  }
}

Look at the r_rows

With the new query everything but the cardid is covered. It takes 1 index access and one table access to find the result.

{
  "query_optimization": {
    "r_total_time_ms": 0.258566593
  },
  "query_block": {
    "select_id": 1,
    "cost": 0.6448338,
    "r_loops": 1,
    "r_total_time_ms": 0.712225566,
    "temporary_table": {
      "nested_loop": [
        {
          "table": {
            "table_name": "cp",
            "access_type": "range",
            "possible_keys": ["card_name_index", "cards_prop_abid_name_value"],
            "key": "cards_prop_abid_name_value",
            "key_length": "267",
            "used_key_parts": ["addressbookid", "name"],
            "loops": 1,
            "r_loops": 1,
            "rows": 638,
            "r_index_rows": 638,
            "r_rows": 1,
            "cost": 0.6448338,
            "r_table_time_ms": 0.672583116,
            "r_other_time_ms": 0.029195924,
            "r_engine_stats": {
              "pages_accessed": 6
            },
            "filtered": 100,
            "r_total_filtered": 0.156739812,
            "index_condition": "cp.addressbookid = 18 and cp.`name` = 'FN' and cp.`value` collate utf8mb4_general_ci like '%user0%'",
            "r_icp_filtered": 0.156739812,
            "r_filtered": 100
          }
        }
      ]
    }
  }
}

I have also experimented with cardid but then MariaDB no longer uses the index. This is likely because a custom collation is used on value, so query optimizer sees the index as not worthy.

TODO

• Make the changes
• Test the changes

How to test

1. Check out the branch
2. Run example queries with EXPLAIN

Checklist

• Code is properly formatted
• Sign-off message is added to all commits
• Tests (unit, integration, api and/or acceptance) are included
• Screenshots before/after for front-end changes
• Documentation (manuals or wiki) has been updated or is not required
• Backports requested where applicable (ex: critical bugfixes)

[ChristophWurst]

/backport to stable31

[ChristophWurst]

/backport to stable30

[ChristophWurst]

/backport to stable29

[ChristophWurst]

@ArtificialOwl @skjnldsv @sorbaugh 🏓 for review