Summary:
Ref T11404. Depends on D16350.
Currently, custom fields can issue "N+1" queries in some cases, so querying 100 revisions issues 100 extra queries.
This affects all `*.search` endpoints for objects with custom fields, and some older endpoints (notably `differential.query`).
This change bulk loads "normal" custom fields, which gets rid of some of these queries. Instead of loading fields for each object, we build a big list of all fields and load them all at once.
The next change will tackle the remaining inefficient edge queries.
Test Plan:
- Configured a custom field with normal database storage in Differential.
- Ran `differential.query`, looking at custom fields in results for correctness.
- Ran `differential.revision.search`, looking at custom fields in results for correctness.
- In both cases, observed queries drop from `3N` to `2N` (all the "normal" custom field stuff got bulk loaded).
Reviewers: yelirekim, chad
Reviewed By: chad
Maniphest Tasks: T11404
Differential Revision: https://secure.phabricator.com/D16351
Summary:
Ref T1191. For most text columns, we either don't care if "a" and "A" are the same, or we expect them to be different (for example: keys, domains, secrets, etc). Default text columns to the `_bin` collation so they are compared by strict character value. This is safer in cases where we aren't sure.
For some text columns, we allow the user to sort by the column in the UI (like Maniphest task titles) or we do care that "A" and "a" are the same (for example: project names). Introduce a new class of virtual data types, the "sort..." types, to cover these columns. These are like the "text..." types but use sorting collations which treat "A" and "a" the same.
Test Plan:
- Made an effort to identify all columns where the UI relies on database collation.
- Ran `bin/storage adjust` and cleared all warnings.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: beng, epriestley
Maniphest Tasks: T1191
Differential Revision: https://secure.phabricator.com/D10602
Summary:
Ref T1191.
- Adds support for custom fields.
- Adds support for partial indexes (indexes on a prefix of a column).
- Drops old auxiliary storage table: this was moved to custom field storage about a year ago.
- Drops old project table: this was moved to edges about two months ago.
Test Plan:
- Viewed web UI, saw fewer issues.
- Used `grep` to verify no readers/writers for storage or project table.
Reviewers: btrahan
Reviewed By: btrahan
Subscribers: epriestley
Maniphest Tasks: T1191
Differential Revision: https://secure.phabricator.com/D10526
Summary:
Ref T2625. Ref T3794. Ref T418. Ref T1703.
This is a more general version of D5278. It expands CustomField support to include real integration with ApplicationSearch.
Broadly, custom fields may elect to:
- build indicies when objects are updated;
- populate ApplicationSearch forms with new controls;
- read inputs entered into those controls out of the request; and
- apply constraints to search queries.
Some utility/helper stuff is provided to make this easier. This part could be cleaner, but seems reasonable for a first cut. In particular, the Query and SearchEngine must manually call all the hooks right now instead of everything happening magically. I think that's fine for the moment; they're pretty easy to get right.
Test Plan:
I added a new searchable "Company" field to People:
{F58229}
This also cleaned up the disable/reorder view a little bit:
{F58230}
As it did before, this field appears on the edit screen:
{F58231}
However, because it has `search`, it also appears on the search screen:
{F58232}
When queried, it returns the expected results:
{F58233}
And the actually good bit of all this is that the query can take advantage of indexes:
mysql> explain SELECT * FROM `user` user JOIN `user_customfieldstringindex` `appsearch_0` ON `appsearch_0`.objectPHID = user.phid AND `appsearch_0`.indexKey = 'mk3Ndy476ge6' AND `appsearch_0`.indexValue IN ('phacility') ORDER BY user.id DESC LIMIT 101;
+----+-------------+-------------+--------+-------------------+----------+---------+------------------------------------------+------+----------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------------+--------+-------------------+----------+---------+------------------------------------------+------+----------------------------------------------+
| 1 | SIMPLE | appsearch_0 | ref | key_join,key_find | key_find | 232 | const,const | 1 | Using where; Using temporary; Using filesort |
| 1 | SIMPLE | user | eq_ref | phid | phid | 194 | phabricator2_user.appsearch_0.objectPHID | 1 | |
+----+-------------+-------------+--------+-------------------+----------+---------+------------------------------------------+------+----------------------------------------------+
2 rows in set (0.00 sec)
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T418, T1703, T2625, T3794
Differential Revision: https://secure.phabricator.com/D6992
Summary:
Ref T1703. We have currently have two custom field implementations (Maniphest, Differential) and are about to add a third (User, see D6122). I'd like to generalize custom fields before doing a third implementation, so we don't back ourselves into the ApplicationTransactions corner we have with Maniphest/Differential/Audit.
For the most part, the existing custom fields work well and can be directly generalized. There are three specific things I want to improve, though:
- Integration with ApplicationSearch: Custom fields aren't indexable. ApplicationSearch is now online and seems stable and good. D5278 provides a template for a backend which can integrate with ApplicationSearch, and ApplicationSearch solves many of the other UI problems implied by exposing custom fields into search (principally, giant pages full of query fields). Generally, I want to provide stronger builtin integration between custom fields and ApplicationSearch.
- Integration with ApplicationTransactions: Likewise, custom fields should support more native integrations with ApplicationTransactions, which are also online and seem stable and well designed.
- Selection and sorting: Selecting and sorting custom fields is a huge mess right now. I want to move this into config now that we have the UI to support it, and move away from requiring users to subclass a ton of stuff just to add a field.
For ApplicationSearch, I've adopted and generalized D5278.
For ApplicationTransactions, I haven't made any specific affordances yet.
For selection and sorting, I've partially implemented config-based selection and sorting. It will work like this:
- We add a new configuration value, like `differential.fields`. In the UI, this is a draggable list of supported fields. Fields can be reordered, and most fields can be disabled.
- We load every avialable field to populate this list. New fields will appear at the bottom.
- There are two downsides to this approach:
- If we add fields in the upstream at a later date, they will appear at the end of the list if an install has customized list order or disabled fields, even if we insert them elsewhere in the upstream.
- If we reorder fields in the upstream, the reordering will not be reflected in install which have customized the order/availability.
- I think these are both acceptable costs. We only incur them if an admin edits this config, which implies they'll know how to fix it if they want to.
- We can fix both of these problems with a straightforward configuration migration if we want to bother.
- There are numerous upsides to this approach:
- We can delete a bunch of code and replace it with simple configuration.
- In general, we don't need the "selector" classes anymore.
- Users can enable available-but-disabled fields with one click.
- Users can add fields by putting their implementations in `src/extensions/` with zero subclassing or libphutil stuff.
- Generally, it's super easy for users to understand.
This doesn't actually do anything yet and will probably see some adjustments before anything starts running it.
Test Plan: Static checks only, this code isn't reachable yet.
Reviewers: chad, seporaitis
Reviewed By: chad
CC: aran
Maniphest Tasks: T1703
Differential Revision: https://secure.phabricator.com/D6147
