getting odd features

[bodritto]

queries like: http://eu.barzer.net/query/json? key=aRLsIvszISAReCoS6ktgviZxN0YlRpbs6DKH7vro& route=doc.features& zurch=0& query=2.DOC_NUM

all docuemtns having doc_id=2.xx probably got all features in index

query exz: http://eu.barzer.net/query/json?key=aRLsIvszISAReCoS6ktgviZxN0YlRpbs6DKH7vro&route=doc.features&zurch=0&query=2.1

[0xd34df00d]

Technically speaking, that's by design. Uniqueness is not limited anyhow, and since all features get into the index, you actually get all the features. We need to have a way to pass uniqueness/gramsize filter to the getter.

[0xd34df00d]

q parameter should be used instead of query

[bodritto]

actually not:

http://eu.barzer.net/query/json?key=aRLsIvszISAReCoS6ktgviZxN0YlRpbs6DKH7vro&route=doc.features&zurch=0&q=2.1

[0xd34df00d]

Because /zurch path should be used :)

http://localhost:8080/zurch?u=1000200&route=doc.features&q=2.1 worksforme, so do http://localhost:8080/zurch?u=1000200&route=doc.features&q=2.1&uniq=2 or http://localhost:8080/zurch?u=1000200&route=doc.features&q=2.1&uniq=2&mingramsize=2&maxgramsize=3

For some reason the corresponding queries on eu.barzer.net (like http://eu.barzer.net/zurch?key=aRLsIvszISAReCoS6ktgviZxN0YlRpbs6DKH7vro&route=doc.features&zurch=0&q=2.1 ) are rendered with the 502 error. @pltr , could you please take a look?

[pltr]

/zurch on eu. only works for debugging and only when you start barzer as http there manually (it isn't started by default). and of course key parameter will do nothing for zurc anyway, it's a proxy parameter

[0xd34df00d]

Thanks, that was already clarified with Igor this day.

[0xd34df00d]

Checked with query tag/stuff, seems to return only required features as well:

cmd>sr u="1000200" route="doc.features" uniq="2" maxgramsize="1" mingramsize="1" zurch="0"
2.1
QUERY:<query u="1000200" route="doc.features" uniq="2" maxgramsize="1" mingramsize="1" zurch="0">2.1</query>
{
"ents" : [
],
"entless" : [
    {
        "size": 1,
        "uniq": 2,
        "grams": [ {"text": "указан" } ]
        ,"poslist": [ { "pos": 7, "len": 12 } ]
    }
,    {
        "size": 1,
        "uniq": 2,
        "grams": [ {"text": "верн" } ]
        ,"poslist": [ { "pos": 18, "len": 14 } ]
    }
,    {
        "size": 1,
        "uniq": 2,
        "grams": [ {"text": "был" } ]
        ,"poslist": [ { "pos": 0, "len": 6 } ]
    }
]
}

All other docs I've (randomly) checked return similarly unique features.

[bodritto]

so now if doc_id in query is empty or non-existing, all features in index returned

i think it would be better if this like queries have empty feature list, and all features if needed should be returned by special flag or pseudo doc id, smth like "_all"

@barzerman what do u think?

[0xd34df00d]

Now an error is returned if document ID is unknown:

QUERY:<query u="1000200" ret="json" route="doc.features" zurch="0">231321</query>
{ "error": "unknown document ID is passed", "id": "231321" }

I've also added validation for input parameters to avoid confusion like this:

QUERY:<query u="1000200" ret="json" route="doc.features" zurch="0" maxgramsize="shitfuck">2.1</query>
{ "error": "invalid options passed" }

We could also later add some kind of error codes to simplify client-side error handling.

The shit is in master. I think the issue can be closed. Reopen if you feel like.

[0xd34df00d]

And, well, fetching all features for all docs is done by an empty query or a query consisting of a single space sign.