Can't create text index

[etuini]

I used the docker-compose example with couchdb and dreyfus images and couchdb works fine, but I can't create a text index. Is there something else I need to do to get text indexes to work?

Thanks.

Brandon

[dmunch]

Hi,

may I ask how you're trying to create a text index?

Note that there isn't a dedicated UI to it like the one for Cloudant. For creating a text-index in this setup you have to manually create design documents like explained here https://docs.cloudant.com/search.html

Thanks for using this repository!
Daniel

[etuini]

Thank you for setting this up.

I was trying to index all fields using the design doc below.

{
"_id": "_design/mango",
"_rev": "1-cf69f994141ea2e0f8c8cf5990b3c9a8",
"language": "query",
"indexes": {
"fulltext": {
"index": {
"default_analyzer": "keyword",
"default_field": {},
"selector": {},
"fields": "all_fields",
"index_array_lengths": true
},
"analyzer": {
"name": "perfield",
"default": "keyword",
"fields": {
"$default": "standard"
}
}
}
}
}

[dmunch]

For creating full-text Mango Indexes you have to explicitly specify the index type, e.g. "type": "text" - A more detailed description can be found here for example: https://cloudant.com/blog/mango-json-vs-text-indexes/

[etuini]

Do I have to explicitly define an index function, like this?
{
"_id": "_design/search_example",
"indexes": {
"animals": {
"index": "function(doc){ ... }"
}
}
}

I'm confused about when to use examples like this:
{
"index":{
},
"name" : "array-index",
"type" : "text"
}

Is the the second example a shortcut that Fauxton translates into index functions?

[dmunch]

You have to differentiate the two approaches: The first approach, where you define an index function is what this documentation is about: https://docs.cloudant.com/search.html - That's where you have the most control and where you write your search queries in Lucene query syntax.

The second approach uses Mango which is a query language inspired by the MongoDb query language. Mango has its own indexes, and one type of index is the full-text index powered by Lucene. Internally what happens is that the Mango Query you're writing get's translated in to a Lucene query and send through the same search pipeline as for the first approach.

If you're confused, that's okay, documentation still seems to be a little sparse on the differences. If in doubt, I found it easier reasoning about it by only using the first approach in the beginning. The way you define the index feels pretty CouchDb-ish, it is very similar to map function in MapReduce views.

[etuini]

I tried to create a mango index of all fields the text below, but it doesn't work. In Fauxton, I get a message that says "! text".
{
"type": "text",
"index": {}
}

I tried another docker image (homerjam/couchdb-cloudant-search) and the index above worked. That's why I was wondering if I need to do something else with your docker image to get text indexes to work.

Thanks for your help on this.

[dmunch]

Ok I'll check tomorrow once I got access to a bigger machine. I didn't put too much focus on Mango so maybe something slipped through.

[dmunch]

So I checked on this and it indeed exhibits a weird behavior. I can't create a Mango based text index up until I executed at least one single request on the index function based search endpoint (i.e. creating at least one index function based search index and querying it).

I could imagine that it's related to how I split up the couchdb container from the clouseau container, because now technically they don't run on the same host any longer and communicate via erlang long node names. The other image you cited doesn't do that, it puts couchdb and clouseau into one single container.

Unfortunately I don't currently have time to look into this, but I'll leave this issue open for future reference.