Now largely automated:
VARS=$(grep -o "'[A-Za-z0-9_]*'" includes/MainConfigNames.php | \
tr "\n" '|' | sed "s/|$/\n/;s/'//g")
sed -i -E "s/'($VARS)'/MainConfigNames::\1/g" \
$(grep -ERIl "'($VARS)'" includes/)
Then git add -p with lots of error-prone manual checking. Then
semi-manually add all the necessary "use" lines:
vim $(grep -L 'use MediaWiki\\MainConfigNames;' \
$(git diff --cached --name-only --diff-filter=M HEAD^))
I didn't bother fixing lines that were over 100 characters unless they
were over 120 and triggered phpcs.
Bug: T305805
Change-Id: I74e0ab511abecb276717ad4276a124760a268147
- Added a test where ParserOutput objects with CacheTime
properties set are unserialized from previous versions.
- Generate new serialization tests for 1.38
Now all serialization in production is JSON, so changing
property visibility shouldn't affect ParserCache.
Bug: T263851
Depends-On: I283340ff559420ceee8f286ba3ef202c01206a23
Change-Id: I70d6feb1c995a0a0f763b21261141ae8ee6dc570
== History of WikiPage::triggerOpportunisticLinksUpdate ==
* 2007 (r19095; T10575; b3a8d488a8)
Introduces the "cascading protection" feature.
This commit added code to Article.php, in a conditional branch
where we encountered a ParserCache "miss" and thus have done a
fresh parse. The code in question would query which templates
we ended up using, and if that differed from what the database
said (e.g. stored during the last actual edit or links update),
then a new LinksUpdate is ad-hoc constructed and executed.
I could not find it anywhere explicitly spelled out, but my best
guess is that the reason for this is to make sure that if the page
in question contains wikitext that trancludes a different page based
on the current date and time (such as how most Wikipedia main pages
transclude news information and "Did you know" information based on
dated subpages that are prepared in advance), then we don't just
want to re-render the page after a day has passed, we also want to
re-do the links update to ensure the search index, category links,
and "WhatLinksHere" is correct, and thus by extent, to make sure
that cascading protection from the main page does in fact apply
to the "current" set of subpages and templates actually in-use.
* 2007 (r19227; 0c0c0eff81)
This adds an optimisation to the added logic that limits it to
pages that satisfy `mTitle->areRestrictionsCascading()`.
Thus for most articles, which aren't protected at all, we don't
run LinksUpdate mid-request after a cache miss page view.
Because of this commit, the pre-2007 status quo remained unaltered
and has remains unaltered to this very day: We don't re-index
categories and WhatLinksHere etc, unless an article edit or
propagating template edit takes place.
* 2009 (r52888; 1353a8ba29)
Introduces the PoolCounter feature.
The logic in question moves to Article::doCascadeProtectionUpdates().
* 2015 (Iea952d4d2e66; df5ef8b5d7).
The logic in question is changed, motivated by wanting to avoid
DB writes during page views.
* Instead of executing LinksUpdate mid-request, we now queue a
RefreshLinksJob on the JobQueue, and utilize a newly added
`prioritize => true` parameter.
This commit also introduces a new feature, which is to queue
RefreshLinksJob also for pages that do not have cascading
protection, but that do satisfy a new boolean method
called `$parserOutput->hasDynamicContent()`, which is set when
the Parser encounters TTL-reducing magic words and functions
such as {{CURRENTDAY}} and {{#time}}. For this new case, however,
the `prioritize` parameter is not set, and this feature is disabled
in WMF production (and other farms that enable wgMiserMode).
This commit also renamed doCascadeProtectionUpdates()
to triggerOpportunisticLinksUpdate().
This commit also removed various documentation comments, which
I've partly restored in this patch, the patch you're looking at
now.
== Actual changes ==
* Rename hasDynamicContent() to hasReducedExpiry() and keep the
previous method as a non-deprecated wrapper.
This change is motivated by T280605, in which I intent to make use
of a Parser hook that reduces the cache expiry. There are numerous
extensions in WMF production that already do this, and thus the
assumption that these have "dynamic content" is already false in
some cases. I'm not yet sure how or if to refactor this so to allow
reducing of the TTL *without* causing this side-effect, but as a
first step we can make the method more obvious in its impact
and behaviour.
I've also updated two of the callers that I think will benefit from
this more explicit name and (current) implementation detail.
Bug: T280605
Change-Id: I85bdff7f86911f8ea5b866e3639f08ddd3f3bf6f
CacheTime::mCacheTime can contain the empty string,
which makes it fail on php8
PHP 8 changes the handling for string and int comparison.
These non-numeric strings are no longer silently converted to int 0,
which previously allowed for the comparison to become false without
actually checking against non-integer values.
https://wiki.php.net/rfc/saner-numeric-strings
Bug: T248925
Change-Id: Idecd3a5cd41fc2d54968db4b02787fe2a81af9d2
Fot the sake for forward compatibility, first we need
to remove the check for version in case of a train rollback.
Bug: T269393
Change-Id: Ic7011e20a23e8ad99eb872b050e97baf1ef5c6a2
One major difference with what we've had before is that now we
actually write class names into the serialization - given that
this new mechanism is extencible, we can't establish any kind
of mapping of allowed classes. I do not think it's a problem
though.
Bug: T264394
Change-Id: Ia152f3b76b967aabde2d8a182e3aec7d3002e5ea
CacheTime::mUsedOptions and ParserOutput::mAccessedOptions
do exactly the same thing and has to be merged into a single property.
This patch adds forward-compatibility and needs to be deployed
at least one train before the patch which actually merges the properties.
Change-Id: Ic9d71a443994e2545ebf2a826b9155c82961cb88
This adds JSON serialization and deserialization capabilities
to CacheTime and ParserOutput.
NOTE: JSON serialization is disabled for now. Merging this patch
should not change behavior in production.
Bug: T263579
Change-Id: I18187e8bce573d21f6f1bd29106e07c63a6d2f4d
This reverts commit a4dc6d82af.
I've reverted the merged patch since I didn't do enough testing
on serialized/reserialized ParserOutput and CacheTime. Now I'm
confident serialization/deserialization works.
Changes since original reverted version:
- Use __get/__set instead of DeprecationHelper in order to
avoid $deprecateProperties array to be serialized.
- Add test for old format serialization new format deserialization.
Change-Id: Ic911c2724ad709931d3316e609781fb89b5b7b28
I've been trying to find how -1 can actually get to
the cacheTime and could find any usages in core or
extensions. The proper way of making ParserOutput
uncacheable is to use updateExpiry( 0 ), that is
documented and has some warning about performance costs.
Setting the cacheTime to -1 seems to be an undocumented
feature (it is mentioned in comments as undocumented)
that doesn't seem to be used anywhere anymore.
I intend to remove it entirely after the deprecation is
deployed and we so not see any warnings for a few weeks.
Change-Id: Ie90b7e4a21faae726940fa9082f2e6a6ea8df613
This reverts commit 799c10b7eb.
Reason for revert: Didn't test how this would work with deserializing stored ParserOutput.
Change-Id: I4221bc26282f3b4bd044f0ab50d00e77eb57ede0
* In preparation for ParserCache/Parsoid integration, it's nice to
do some cleanups. Will untie our hands a bit more.
* Verified no usages in extensions deployed at wikimedia, other then
Flow, fixed in the dependent patch.
Change-Id: Idd78413a36887e2ff5c902d410e55691cafb736b
Previously, getCacheTime would default to the current time, potentially
causing the return value to change over subsequent calls. With this change,
the value is determined on the first call, and then remembered for subsequent
calls.
Bug: T205464
Change-Id: If240161c71d523ad5b0d33b9378950e0bebceb6e
Few maintained extensions still rely on this and it is
bad practice to use this for handling cache correctness.
Change-Id: I2de481198bbff5c4f3dd81fc6d1b137e4c37b93f
* Setting mCacheTime to -1 is for old callers that
only check getCacheTime() instead of getCacheExpiry().
Most of them are already broken (WikiLog/SemanticForms) as
they check for -1 which is in fact never returned
due to the TS_MW conversion in Parser::getCacheTime.
* By using -1, the value of page_links_updated can end up
as 1969, which is confusing and broken.
Change-Id: I8809a4258eacff05992a2c27ade7f6a0c1731c51
This doesn't fix all the files under includes/parser -
some of them deserve their own patches.
Bug: T102614
Change-Id: I2fcbc19ee337e1b7db4635b5e5f324c651b4d144
* This should not happen as doEditContent() saves the parser cache,
so only the rare casing if incompatible options should have misses
* The bug could also cause post-save misses with edit stashing
* Avoid the second page parse post-redirect by making sure cache
timestamps match up instead of calling time() at several points
* Likewise for null edits, which used a different code path
* Removed redundant purge in onArticleCreate() as the new row sets _touched
* Removed pointless purge in onArticleDelete() as there is no row to update
(the method no-ops in that case to avoid contention already)
Change-Id: I178fe334a3f8691ffd9452bec30561a0c5d37c6c
Remaining are the classes containing underscores and possibly a few other
issues that will be addressed soonish.
Change-Id: Icf56374c71afc134420ebbcfecf12dcb29dc9564
Swapped some "$var type" to "type $var" or added missing types
before the $var. Changed some other types to match the more common
spelling. Makes beginning of some text in captial.
Also added some missing @param.
Change-Id: I49f8f48b521878de7abd9cc40efdeff6cf9a37e0
One theory for what's behind bug 46014 is that the vandal submits the
edit, then someone (maybe the vandal) gets into the branch of
Article::view that uses PoolWorkArticleView, then ClueBot comes along
and reverts before the PoolWorkArticleView actually executes. Once that
PoolWorkArticleView actually does execute, it overwrites the parser
cache entry from ClueBot's revert with the one from the old edit.
To detect this sort of thing, let's include the revision id in the
parser cache entry and consider it expired if that doesn't match. Which
makes sense to do anyway.
And for good measure, let's have PoolWorkArticleView not save to the
parser cache if !$isCurrent.
Bug: 46014
Change-Id: Ifcc4d2f67f3b77f990eb2fa45417a25bd6c7b790
* This reuses the parser cache in some cases when possible
* Clarified the return value of CacheTime::getCacheTime()
* A few documentation tweaks
Change-Id: I80b7c6404b3f8c48b53c3bba96115dbf94d80873
Doxygen expects parameter types to come before the
parameter name in @param tags. Used a quick regex
to switch everything around where possible. This
only fixes cases where a primitve variable (or a
primitive followed by other types) is the variable
type. Other cases will need to be fixed manually.
Change-Id: Ic59fd20856eb0489d70f3469a56ebce0efb3db13
