[CALCITE-175] Modify Schema SPI to allow caching

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Labels: github-import
• Assignee: Unassigned
• Reporter: GitHub Import
• Created: 18/Jun/14 08:26
• Updated: 18/Jun/14 08:27
• Resolved: 18/Jun/14 08:27

Description

The Schema SPI provides tables, table functions, and sub-schemas via getTableNames, getTable, getTableFunctionNames, getTableFunctions, getSubSchemaNames, getSubSchema. And AbstractSchema additionally provides getTableMap(), getTableFunctionMap(), getSubSchemaMap().

However, we do not specify what should happen if the schema changes; for example, what `CsvSchema` should do it someone adds a .csv file to the directory. Should it re-derive every time? Should optiq assume that the map never changes, and therefore can build a cache?

This is especially relevant when handling case-insensitive name resolution. The above methods are (and will remain) case-sensitive. So the only way to find out whether there is a case-insensitive match for a table (or table-function or sub-schema) is to read all keys in the map. We would like to build a cache, but we need to make sure that the cache does not overstay its welcome.

We could add the following method to `Schema`:

```java
/** Returns whether the contents of the schema (tables, table functions and

• sub-schemas) have changed since a given system time. If this method
• returns true, Optiq will rebuild the cache that it built at {@code time}

  .

• The default implementation in {@link AbstractSchema}

  always returns

• {@code false}

  . */
  boolean contentsHaveChangedSince(long time);
  ```

and a default implementation in `AbstractSchema`.

---------------- Imported from GitHub ----------------
Url: https://github.com/julianhyde/optiq/issues/175
Created by: julianhyde
Labels: enhancement,
Created at: Mon Mar 10 23:26:54 CET 2014
State: closed