Elasticsearch: TotalHitCountCollector message for top hits aggregation can be improved

[ppf2]

If you attempt to specify a size of 0 or -1 for a top hits aggregation, you will get this error:

IllegalArgumentException[numHits must be > 0; please use TotalHitCountCollector if you just need the total hit count

Probably not a good idea to return all hits to begin with, but it will be nice to return a more intuitive message if one does attempt to set this to 0 or -1, etc..

[clintongormley]

@ppf2 can you suggest one?

[martijnvg]

I think if size is set to 0 we should just return a hits that only contains a total count and if a negative value is specified we should fail with a parse exception or something like that.

[ppf2]

I agree, it makes sense to allow size=0, and a parse error for size < 0 is helpful. Maybe indicate in the error message what is the range of size that is allowed (or even warn users about setting size to something very large like integer max, etc..)

[jpountz]

When counting when the size is 0 is what we do for regular search operations. However for aggregations, you already know the match count anyway thanks to the parent aggregation, to I think it's right to raise an error is size is less than or equal to 0 as it's almost certainly a user error.

[linkwoman]

This would be useful to have size:0 allowed. Why? Because I want the max_score for each bucket which I can only get when I return at least 1 top_hit. I do not want to see the top_hit - I just want a report of all buckets, doc counts in each bucket, and max score. Is there another way to get max_score of each bucket? Also, I need min_score. Is there a way to get that @clintongormley ?
thanks.

[jpountz]

@linkwoman you could use a max aggregation on a script that just consists of _score for that. This would also be much more efficient.

[elasticmachine]

Pinging @elastic/es-analytics-geo

[rmadisonhaynie]

When counting when the size is 0 is what we do for regular search operations. However for aggregations, you already know the match count anyway thanks to the parent aggregation, to I think it's right to raise an error is size is less than or equal to 0 as it's almost certainly a user error.

@jpountz Is it possible to reference the match count in the same query or would I need to perform 2 requests? I would like to get all my hits, not just the top N

[OneCodeMonkey]

Probably because you assign an negative value to the from parameter, coz I 'm encounter the same mistake...

[imotov]

I cannot reproduce it in master anymore. I am getting the following error message:

{
  "error" : {
    "root_cause" : [
      {
        "type" : "illegal_argument_exception",
        "reason" : "[from] must be greater than or equal to 0. Found [-1] in [top_tag_hits]"
      }
    ],
    "type" : "illegal_argument_exception",
    "reason" : "[from] must be greater than or equal to 0. Found [-1] in [top_tag_hits]"
  },
  "status" : 400
}

If it is still reproducible somewhere, please free to reopen.