search

gbv / jskos-server

Web service to access JSKOS data
https://coli-conc.gbv.de/api/
MIT License
6 stars 4 forks source link

Certain search queries currently do not work properly #138

Closed stefandesu closed 3 years ago

stefandesu commented 3 years ago

In v1.2.3, I added escaping to search queries for /search and /suggest because certain characters used in regular expressions would cause the request to crash and return an error 500. However, now it won't include those characters at all in the search.

Example of failing query: https://coli-conc.gbv.de/api/suggest?voc=http:%2F%2Fdewey.info%2Fscheme%2Fedition%2Fe23%2F&search=612.112

Search query before escaping: 612.112 Search query after escaping: 612\.112 MongoDB search query: {"$and":[{"$or":[{"_id":"612\\.112"},{"$text":{"$search":"\"612\\.112\""}},{"_keywordsLabels":{"$regex":"^612\\.112"}}]},{"$or":[{"inScheme.uri":"http://bartoc.org/en/node/241"},{"inScheme.uri":"http://www.wikidata.org/entity/Q48460"},{"inScheme.uri":"http://dewey.info/scheme/edition/e23/"}]}]}

As you can see, for some reason after assembling the MongoDB search query, the \ that's supposed to escape the . gets escaped itself (despite not using any escaping methods). When the escaping step is skipped, then it'll use 612.112 where the . is interpreted as part of a regular expression.

stefandesu commented 3 years ago

The problem was that we were providing a string as a regular expression for the Mongo query, and the / that was supposed to escape the . was then interpreted as a normal character instead of an escaping character. The solution is to use RegExp objects instead of strings.