This is expected behavior with _all field. For demonstration I extended your gist a bit.
https://gist.github.com/jprante/9891706 Some hints: - custom tokenizer should be used in a field that is configured in a mapping - always set both search and index analyzer for a field - avoid setting up a custom tokenizer for _all when including more than one field to _all (which is the default). This will give unpredictable results because tokens from many fields are merged into _all. In edge cases, when a field is first for example, you may be able to produce a hit. But this is pure accidentally. - when searching with q parameter, do not forget to specify field name Jörg On Mon, Mar 31, 2014 at 2:23 PM, Huy Phan <[email protected]> wrote: > Hi Luca, > > The configuration index.analysis.analyzer.default_index is already set so > I don't think there's a need to specify my mappings since I actually want > to use the comma analyzer for all the fields. And from what I understand, > that default_index is also applied to _all field. > As you could see in my gist, I also overrode the "standard" analyzer since > I doubted something went wrong with defaul_index. > > You may ask about the default_search configuration, my query "123456" is > rather simple so I don't think the default analyzer would make any changes > on it (and yes, I did verify that using the Analyzer API). > > Even if there's something wrong with my settings, that still doesn't > clearly explain why I got the result with the second document but not with > the first one. > > > On Monday, 31 March 2014 19:45:42 UTC+8, Luca Cavanna wrote: >> >> As far as I can see from your recreation you only create the analyzer but >> don't associate it to your fields by specifying your mappings. Also, when >> you query you don't soecify the field you want to query, thus you are using >> the _all which has its own analyzer, which means that even if you had >> specified the proper mappings the query would execute against a different >> field with a different analyzer. >> >> On Monday, March 31, 2014 12:12:37 PM UTC+2, Huy Phan wrote: >>> >>> Hi all, >>> >>> I bumped into this weird behavior of Elasticsearch: https://gist. >>> github.com/huyphan/9888959<https://www.google.com/url?q=https%3A%2F%2Fgist.github.com%2Fhuyphan%2F9888959&sa=D&sntz=1&usg=AFQjCNH4SNtSUHvK2yfyGrFL2mqfyD-vIQ> >>> >>> Basically what I did is to create a comma analyzer and and use it as the >>> default one. Then I indexed this document >>> >>> { >>> "random_string" : "ABC,XYZ", >>> "random_number" : "123456,7890123", >>> "random_email" : "[email protected],[email protected]" >>> } >>> >>> >>> Then search for it with query "123456", I got no hit. However if I did >>> everything from scratch and indexed a slightly different document (it's >>> actually the same doc with first field removed): >>> >>> { >>> "random_number" : "123456,7890123", >>> "random_email" : "[email protected],[email protected]" >>> } >>> >>> >>> The same old query did give me the result. I'm not sure what is the >>> difference between the 2 documents that causes this behavior. >>> >>> >>> -- > You received this message because you are subscribed to the Google Groups > "elasticsearch" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion on the web visit > https://groups.google.com/d/msgid/elasticsearch/9b25e5f4-22a2-48e0-8ab2-4c72f4d8d25e%40googlegroups.com<https://groups.google.com/d/msgid/elasticsearch/9b25e5f4-22a2-48e0-8ab2-4c72f4d8d25e%40googlegroups.com?utm_medium=email&utm_source=footer> > . > > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "elasticsearch" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAKdsXoHOz2sLRPdqSWY1B3m7tuaXVc%2BRjCEdhProONDwXc3TUg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
