Re: I can't find anything after hypens or underscores

[Alessandro Bonfanti]

Il 26/01/2015 16:37, Alessandro Bonfanti ha scritto: Il 21/01/2015 11:43, Alessandro Bonfanti ha scritto: Il 02/12/2014 09:21, Alessandro Bonfanti ha scritto: Il 12/11/2014 17:43, Alessandro Bonfanti ha scritto: Il 12/11/2014 17:20, Nikolas Everett ha scritto: On Wed, Nov 12, 2014 at 11:13 AM, Alessandro Bonfanti <bnf....@gmail.com> wrote: Il 12/11/2014 15:25, Nikolas Everett ha scritto: On Wed, Nov 12, 2014 at 8:15 AM, Alessandro Bonfanti <bnf....@gmail.com> wrote: Hi, I'm very newbie on ElasticSearch. I'm try to indexing a set of biological data. There are some fields like 'gene_id' or 'gene_shortname' that should be processed as literal strings. When I try to search for 'ZNF6092' in a field filled with 'linc-ZNF6092-6', I can't find anything. When I search for 'linc' I find correct document elsewhere. It seems that this is a problem with ES analyzer, but I tried to set it for do not analyze fields, but it seems that nothing changes. I try with: curl -XPOST 'localhost:9200/a3' -d @tracking_map.json where tracking_map.json is {   "mappings": {     "tracking": {       "properties": {         "tracking_id" : {           "type": "string",           "index":"not_analyzed"         },         "nearest_ref_id" : {           "type": "string",           "index":"not_analyzed"         },         "gene_id" : {           "type": "string",           "index":"not_analyzed"         },         "gene_short_name" : {           "type": "string",           "index":"not_analyzed"         }       }     }   } } And then re-indexing of all documents. I failed, but where? Thanks in advance, Alessandro Its an analyzer problem, certainly.  You've turned off analyzers with "index":"not_analazyed".  What you probably want is for the gene_short_name to be analyzed so that dashes are considered "word separators".  If you do that you can find linc-ZNF6092-6 by performing a simple_query_string (or match) search for <code>ZNF6092</code> or <code>ZNF6092 6</code> or <code>6</code> or <code>linc</code>.   Have a look at http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/analysis-pattern-tokenizer.html and go from there.  You may also want to use a lowercase filter so you can search for <code>znf6092</code> and still find it. This is a good read on how to change the mapping as well: http://www.elasticsearch.org/blog/changing-mapping-with-zero-downtime/ even if you don't need all the information in there it is nice to know. Nik -- You received this message because you are subscribed to a topic in the Google Groups "elasticsearch" group. To unsubscribe from this topic, visit https://groups.google.com/d/topic/elasticsearch/Y6I2qNZxR-s/unsubscribe. To unsubscribe from this group and all its topics, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAPmjWd06sKTVS6JC8q7x7R37gUEnsHEiuar0-yy_ZdOJQhKYzQ%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout. Very thanks for your answer, What I want is that ES store fields as literals, so I should find ZNF6092 with a wilcard search (*ZNF6092* for example). I tried set "pattern" to "*" for testing (* isn't in gene_shortname, so I suppose that entire string is stored. But anyway I still find nothing. You'd have to post your queries for me to help more but in general if best to analyze the content up front and perform basic match queries without wildcards than it is to search with wildcards.  Wildcards are way way way slower. Nik -- You received this message because you are subscribed to a topic in the Google Groups "elasticsearch" group. To unsubscribe from this topic, visit https://groups.google.com/d/topic/elasticsearch/Y6I2qNZxR-s/unsubscribe. To unsubscribe from this group and all its topics, send an email to elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAPmjWd0itbdHQ-maOuOmrrYf2QCqMFORTG21QpFHOCrp9E0rmg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout. This is my query (in Ruby): @client.search index: @index, body: {query: {wildcard: {_all: query_text}}} Variables' name should be auto-explicative of its content. I read that wildcards are slower, if you have a more clean solution (I need anyway that I still can search for "linc-ZNF6092" in addiction for "ZNF6092") it will be very welcome. I have tried a lot of attempts, but the problem still resist. Maybe could it be caused by another setting than analyzer? Definitely, I need a step-to-step method for disabling the analyzer or set it to 'keyword' on all fields of an index. I tried a lot of attempts but no-one seems to work. This situation cause me much problems, I need that ES do not tokenize my literal strings, why there isn't a clear method to switch of it? Thanks everyones. OK, after a lot of attempts I can finally set analyezer to 'keyword' for default. I do this with: @es_client.indices.create index: "test", body: { "index" => { "analysis" => { "analyzer" => { "default" => { "type" => "keyword" }}}}} Now I have solved some problems, I finally can do exact matching stuff with 'term' query, for example on a path '/home/data/foo.bar' or on a gene-id 'ENSG00000186092'. The bad things are that problems with 'query_string' even worsen. It seems that query_string can't work with not analyzed fields. If I try a trivial: @es_client.search index: "test", body: {"query" => { "query_string" => { "query" => "ENSG00000186092" }}} Nothing works (0 results found). Text hasn't spaces or other special characters that could create problems with tokenization. So what's the problem? Can a solution be the use of a 'fake' pattern tokenizer with pattern "$^" (this should create a non-matchable pattern, with result alike the 'keyword' analyzer)? Any other idea will be very appreciated. Problems with search derived probably by the fact that query_string automatically make lovercased all words. It's behavior caused by 'lowercase' filter automatically inserted. I can't find on the web any examples about setting of analyzers/tokenizers/filters via ruby APIs. The only one that seems to work well is the pulled over method for set the default analyzer when a new index is created. Any suggestion? I need valid method for set them in custom fields/searches etc.  -- 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 elasticsearch+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/54C8B2C6.90300%40gmail.com. For more options, visit https://groups.google.com/d/optout.