[
https://issues.apache.org/jira/browse/SOLR-7913?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17267477#comment-17267477
]
Isabelle Giguere edited comment on SOLR-7913 at 1/18/21, 7:42 PM:
------------------------------------------------------------------
MLT Query Parser was originally implemented to allow field queries (i.e.:
myField:some text)
https://issues.apache.org/jira/browse/SOLR-6248
Read specifically the discussion between Steve Molloy, Vitaliy Zhovtyuk and
Anshum Gupta in the first few comments.
By the time the MLT QParser was committed to SVN trunk, the query format was
changed:
https://issues.apache.org/jira/browse/SOLR-6248?focusedCommentId=14189235&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14189235
With this format, the input immediately following the closing curly brace is
assumed to be a docId (whatever field is the unique id key in the schema)
As of now, without any of the patches on this ticket, the first thing that
happens is to look for document by id, and if that fails, throw an exception.
This whole "stream.body" discussion (or monologue) originally started because
of a need to identify a document using a query on any field, not just an id.
Maybe it's time to move away from the idea of "stream.body", and re-implement
support for any field query in MLT QParser.
If the extra tests added in SOLR-7913_negative-tests.patch could produce
results instead of an exception, I don't think anyone would need to use
stream.body with an MLT QParser query.
was (Author: igiguere):
MLT Query Parser was originally implemented to allow field queries (i.e.:
myField:some text)
https://issues.apache.org/jira/browse/SOLR-6248
Read specifically the discussion between Steve Molloy, Vitaliy Zhovtyuk and
Anshum Gupta in the first few comments.
By the time the MLT QParser was committed to SVN trunk, the query format was
changed:
https://issues.apache.org/jira/browse/SOLR-6248?focusedCommentId=14189235&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14189235
With this format, the input immediately following the closing curly brace is
assumed to be a docId (whatever field is the unique id key in the schema)
In CloudMLTQParser, without any of the patches on this ticket, the first thing
that happens is to look for document by id, and if that fails, throw an
exception.
This whole "stream.body" discussion (or monologue) originally started because
of a need to identify a document using a query on any field, not just an id.
Maybe it's time to move away from the idea of "stream.body", and re-implement
support for any field query in MLT QParser.
> Add stream.body support to MLT QParser
> --------------------------------------
>
> Key: SOLR-7913
> URL: https://issues.apache.org/jira/browse/SOLR-7913
> Project: Solr
> Issue Type: Improvement
> Reporter: Anshum Gupta
> Priority: Major
> Attachments: SOLR-7913.patch, SOLR-7913.patch, SOLR-7913.patch,
> SOLR-7913.patch, SOLR-7913_fix-unit-test-setup.patch,
> SOLR-7913_fixTests.patch, SOLR-7913_negative-tests.patch,
> SOLR-7913_tag_7.5.0.patch
>
>
> Continuing from
> https://issues.apache.org/jira/browse/SOLR-7639?focusedCommentId=14601011&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14601011.
> It'd be good to have stream.body be supported by the mlt qparser.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
