[
https://issues.apache.org/jira/browse/SOLR-141?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=12854690#action_12854690
]
Frank Wesemann commented on SOLR-141:
-
{quote}
1) more smarts in SolrCore.getErrorResponseWriter to be able to deal with the
possibility that the ResponseWriter is also an ErrorResponseWriter - that way
people wouldn't have to register a dual purpose impl twice
{quote}
wouldn't it be sufficient to extend the interface QueryResponseWriter with a
writeError(Writer, HttpServletResponse, SolrResponse, SolrRequest) Method?
The DispatchFilter may set the last caught exception on SolrResponse and call
it in the sendError() method.
The ResponseWriter impl may than do something like:
{code}
public void writeError(Writer out, HttpServletResponse res, SolrResponse
solrRes, SolrRequest solrReq) {
Throwable th = solrRes.getException();
res.setStatus(code);
res.setContentType( getContentType() );
myTextResponseWriterImpl.writeInt( code );
myTextResponseWriterImpl.writeStr( th.getMessage );
if ( ! th instanceof QueryResponseWritingExecption ) {
try {
myTextResponseWriterImpl.writeResponse();
} catch ( Exception e) {
myTextResonseWriterIml.writeStr(an additionalError in writing the
result occured: + e.getMessage() );
}
}
//... do what else you need to.
}
{code}
Errors/Exceptions should be formated by ResponseWriter
--
Key: SOLR-141
URL: https://issues.apache.org/jira/browse/SOLR-141
Project: Solr
Issue Type: Wish
Reporter: Hoss Man
Fix For: 1.5
Attachments: solr-exception-writer-solr-1.2.diff,
solr-exception-writer-v2.diff, solr-exception-writer-v3.diff
Whenever possible, the Solr Dispatcher should to let the ResponseWriter
format Exceptions using the format the user expects -- this should in no way
change the fact that Exceptions currently generate non 200 HTTP status codes,
nor should it prevent the Dispatcher from using the exception message as the
HTTP status message -- but clients that want the full details of the error
should be able to parse them in the format they expected based on the request.
this would also give RequestHandlers the oportunity to return partial
results - by adding both whatever results they have to the Response as well
as the Exception they encountered.
situations where this can't happen are obviously:
* Exception thrown by ResponseWriter
* Exception thrown so early in the request thta the DIspather doesn't know
which ResposneWriter the client wanted.
...in those cases, plain text is a wise choice.
thing that would probably need to be done to make this work:
1) if the Dispatcher catches an exception, it should call
SolrQueryResponse.setException, set the HTTP status code/message as it
currently does, but then hand off to the ResponseWriter.
2) Dispatcher needs to check SolrQueryResponse.getException and set the HTTP
status code/message just like if it caught the exception itself.
3) all of the ResponseWriters should start looking at
SolrQueryResponse.getException if they aren't already, and formatting it in a
usefull way.
4) if the ResponseWriter throws an exception, Dispatcher needs to return a
nice plain text error page
extension to this idea... add a new method to ResponseWriter to generate a
generic error message in the appropriate format that Dispatcher can use if
the ResponseWriter throws an exception (as a backup before resorting to plain
text)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
