Trejkaz created LUCENE-7616:
-------------------------------
Summary: QueryNode#toQueryString says it produces a string in the
syntax understood by "the query parser", but cannot possibly know how
Key: LUCENE-7616
URL: https://issues.apache.org/jira/browse/LUCENE-7616
Project: Lucene - Core
Issue Type: Bug
Components: modules/queryparser
Affects Versions: 6.3
Reporter: Trejkaz
(Not an implementation "bug" so much as a design error, but working within the
confines of JIRA here.)
The "flexible" query parser framework allows custom query syntaxes to be
implemented on top of the existing query node, processor and builder classes.
Now, QueryNode has a toQueryString method which ostensibly converts the node
back into a string appropriate for passing back through the parser. However, in
practice, this method is implemented to return a syntax only appropriate for
passing back to StandardQueryParser, *not* the parser you got the node from.
The node itself has no idea what parser it came from, so it makes sense that
this method could never work as currently designed.
I don't really know what the right way to fix this is.
Option A: Make QueryNode aware of which parser it came from, and add methods
into the parser to format queries back into a string, so that this method can
be implemented correctly. Sounds fine, except programmatically creating
QueryNode objects directly becomes a hassle.
Option B: Deprecate toQueryString and introduce a new SyntaxFormatter interface
which converts QueryNode to CharSequence and provide an appropriate
implementation for each existing SyntaxParser. Seems sensible and the most
flexible option, but requires a lot of tiny classes to be implemented.
Are there any other options?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
