Renaming performance indexes on upgrade#342
Conversation
…lineTableUpgrader, OracleMetaDataProvider.
…r, InlineTableUpgrader, OracleMetaDataProvider. #2 Refactor duplicated code.
Changes that apply to more than one file: - Remove ignoredIndexesMap - Remove schemaResources from constructor - Use Table.ignoredIndexes instead of a map for all tables Files: - AbstractSchemaChangeVisitor - DatabaseMetaDataProvider - also add support for ignored indexes at table loading level - GraphBasedUpgradeBuilder - GraphBasedUpgradeSchemaChangeVisitor - InlineTableUpgrader - OracleMetaDataProvider - Upgrade - UpgradeTestHelper Add ignoredIndexes() property: - SchemaUtils.TableBuilder / Impl - Table - TableBean Update tests to pass.
- Re-add schemaResources for constructor of InlineTableUpgrader and GraphBasedUpgradeSchemaChangeVisitor Files: - AbstractSchemaChangeVisitor - GraphBasedUpgradeBuilder - GraphBasedUpgradeSchemaChangeVisitor - InlineTableUpgrader - OracleMetaDataProvider - Upgrade - UpgradeTestHelper Update tests to pass.
… must be called with dataSource argument.
Add ignoredIndexes to AdditionalMetadata. Add getIgnoredIndexes/set and getIgnoredIndexesForTable to UpdateConfigAndContext. Replace Set<String> exclusiveExecutionSteps parameter on GraphBasedUpgradeBuilder by UpdateConfigAndContext. Read ignoredIndexes from AdditionalMetadata on section for healing where we have schemaResources open to avoid opening more than once. Adapt code on GraphBasedUpgradeSchemaChangeVisitor and InlineTableUpgrader to use the UpdateConfigAndContext.getIgnoredIndexesForTable instead. Adapt OracleMetaDataProvider to read ignored indexes into the owned map and expose ignoredIndexes() Adapt PostgreSQLMetaDataProvider to read ignored indexes into the owned map and expose ignoredIndexes() Adapt tests to changes.
|
PostgreSQLMetaDataProvider - way of calculating ignored indexes was changed. the indexes map table name is stored in lower case for consistency. UpgradeConfigAndContext - the setIgnoredIndexes ensures the table name is lowercase in the map stored. Remove unnecessary line endings on TableBean and SchemaUtils.
|
|
||
| private final Supplier<Map<AName, RealName>> allIndexNames = Suppliers.memoize(this::loadAllIndexNames); | ||
| private final Supplier<Map<String, List<Index>>> ignoredIndexes = Suppliers.memoize(this::loadAllIgnoredIndexes); | ||
| private final Map<String, List<Index>> allIgnoredIndexes = Maps.newHashMap(); |
There was a problem hiding this comment.
We should definitely use Suppliers.memoize(this::loadAllIgnoredIndexes) for this.
That loadAllIgnoredIndexes() method can call allIndexNames.get() to populate allIgnoredIndexesTables.
I am not sure I like this side-effect allIgnoredIndexesTables population, but at least it is private to this class.
There was a problem hiding this comment.
that change wont work. I had to change it to load in separate ways. the side effect here loads the ignored indexes on need only after allIndexNames.get() is read (only once) - that method loads the tables and sets allIgnoredIndexesTables.
...
private final Supplier<Map<AName, RealName>> allIndexNames = Suppliers.memoize(this::loadAllIndexNames);
...
public Map<String, List<Index>> ignoredIndexes() {
// make sure allIgnoredIndexesTables is loaded.
allIndexNames.get();
if (allIgnoredIndexes.isEmpty() && !allIgnoredIndexesTables.isEmpty()) {
loadAllIgnoredIndexes();
}
return allIgnoredIndexes;
}
| // make sure allIgnoredIndexesTables is loaded. | ||
| allIndexNames.get(); | ||
| if (allIgnoredIndexes.isEmpty() && !allIgnoredIndexesTables.isEmpty()) { | ||
| loadAllIgnoredIndexes(); |
There was a problem hiding this comment.
There could be schemas with no ignored indexes, re-loading this again and again.
That is currently fairly little extra work; but it does create a potential tech pickle for us in the future.
Suppliers.memoize would solve this problem for you :)
There was a problem hiding this comment.
I don't seem to understand. The condition doesn't load it again and again.
allIndexNames.get() sets allIgnoredIndexTables to be not empty only and only if there are ignored indexes - only once.
On second passes it will never call loadAllIgnoredIndexes() because allIgnoredIndexes now is not empty.
It is a bit obscure, but when you think of it, it works efficiently.
|
3 participants
No description provided.