In support of T191231: RFC: Abstract schemas and schema changes, Math should be migrated to using Abstract Schema
Description
Description
Details
Details
Related Changes in Gerrit:
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Convert Math to abstract schema | mediawiki/extensions/Math | master | +255 -112 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Ladsgroup | T191231 RFC: Abstract schemas and schema changes | |||
| Open | None | T259374 Convert extensions to using abstract schema | |||
| Resolved | None | T261912 Convert WMF Deployed Extensions to Abstract Schema | |||
| Resolved | Physikerwelt | T268561 Convert Math to AbstractSchema |
Event Timeline
Comment Actions
I looked into this task, and writing the json file does not seem to be funny. However, this tool might help https://symfony.com/doc/current/doctrine/reverse_engineering.html
Comment Actions
It's not overly difficult. I'm currently working a script Amir started to make it do most of the migrations https://github.com/Ladsgroup/db-analyzor-tools/blob/master/db_abstractor.py
Comment Actions
That script is now ready and basically gives you everything except comments. The output is
[ { "name": "mathlatexml", "columns": [ { "name": "math_inputhash", "type": "binary", "options": { "notnull": true, "length": 16 } }, { "name": "math_inputtex", "type": "text", "options": { "notnull": true } }, { "name": "math_tex", "type": "text", "options": { "notnull": false } }, { "name": "math_mathml", "type": "mediumtext", "options": { "notnull": false } }, { "name": "math_svg", "type": "text", "options": { "notnull": false } }, { "name": "math_style", "type": "mwtinyint", "options": { "notnull": false } } ], "indexes": [], "pk": [ "math_inputhash" ] } ]
and
[ { "name": "mathoid", "columns": [ { "name": "math_inputhash", "type": "binary", "options": { "notnull": true, "length": 16 } }, { "name": "math_input", "type": "text", "options": { "notnull": true } }, { "name": "math_tex", "type": "text", "options": { "notnull": false } }, { "name": "math_mathml", "type": "text", "options": { "notnull": false } }, { "name": "math_svg", "type": "text", "options": { "notnull": false } }, { "name": "math_style", "type": "mwtinyint", "options": { "notnull": false } }, { "name": "math_input_type", "type": "mwtinyint", "options": { "notnull": false } }, { "name": "math_png", "type": "blob", "options": { "notnull": false, "length": 16777215 } } ], "indexes": [], "pk": [ "math_inputhash" ] } ]
That's a good starting point at least.
Comment Actions
I tried it on the non-wmf deployed mathsearch extension first, as this is less critical. I get the impression that KEY is not recognized as a keyword.
Input
-- -- Used by the math search module to annotate meanings of identifiers. -- CREATE TABLE /*_*/mathidentifier ( identifier varchar(5) NOT NULL, noun varchar(40) NOT NULL, evidence double NOT NULL, pageTitle varchar(255) NOT NULL, pageID int(8) NOT NULL, KEY mathidentifier_key( identifier , pageTitle ) ) /*$wgDBTableOptions*/;
output
[
{
"name": "mathidentifier",
"columns": [
{
"name": "identifier",
"type": "string",
"options": {
"notnull": true,
"length": 5
}
},
{
"name": "noun",
"type": "string",
"options": {
"notnull": true,
"length": 40
}
},
{
"name": "evidence",
"type": "double",
"options": {
"notnull": true
}
},
{
"name": "pagetitle",
"type": "string",
"options": {
"notnull": true,
"length": 255
}
},
{
"name": "pageid",
"type": "integer",
"options": {
"notnull": true,
"length": 8
}
},
{
"name": "key",
"type": "mathidentifier_key",
"options": {
"notnull": false
}
}
],
"indexes": []
}
]Comment Actions
Doing the rest manually is not hard ;)
But it's purposely skipped as the index needs naming, which needs human input - https://github.com/Ladsgroup/db-analyzor-tools/blob/master/db_abstractor.py#L65-L69
Comment Actions
OK. I will figure it out... However, it does not match my KEY mathidentifier_key( identifier , pageTitle ) as the name mathidentifier_key was not expected;-) And that the key name is a type is somehow confusing.
Comment Actions
Yeah, there's no handling for it
"indexes": [ { "name": "mathidentifier_key", "columns": [ "identifier", "pageTitle" ], "unique": false } ],
Comment Actions
Change 652207 had a related patch set uploaded (by Physikerwelt; owner: Physikerwelt):
[mediawiki/extensions/Math@master] Convert Math to abstract schema
Comment Actions
Please read https://www.mediawiki.org/wiki/Manual:Coding_conventions/Database
Naming, prefixes, index names, they all need to be fixed here.
Comment Actions
I did read the document, and still do not understand your plan. Renaming tables or fields is nothing I think would be reasonable to change. Do you think we should have a name for the PK?
Comment Actions
PK doesn't need a name, for PK you can just use the "pk" option. See https://github.com/wikimedia/mediawiki/blob/master/maintenance/tables.json for examples.
Comment Actions
mh.. that is what your script did :-) Maybe you can give me a few hints on gerrit what can / should be improved. However, renaming fields/tables should be done in a separate change I suggest.
Comment Actions
Change 652207 merged by jenkins-bot:
[mediawiki/extensions/Math@master] Convert Math to abstract schema