Adds a setting to disable source output in slowlog

[bookshelfdave]

When logging large documents via slowlog, setting index.slow_logging.source
to false prevents the source from being displayed.

Closes #4485

I don't see any tests for slowlog, so I used these steps to manually test:

curl -XPUT 'http://localhost:9200/bar'

# don't log source
curl -XPUT 'localhost:9200/bar/_settings' -d '
{
    "index" : {
        "slow_logging.source" : false
    }
}'

# bump the index threshold to 0s
curl -XPUT 'http://localhost:9200/bar/_settings' -d '
{
    "index.search.slowlog.threshold.query.warn" : "10s", 
    "index.search.slowlog.threshold.fetch.debug": "500ms", 
    "index.indexing.slowlog.threshold.index.info": "0ms" 
}'



curl -XPUT 'http://localhost:9200/bar/tweet/2' -d '
{
    "user": "kimchy",
    "postDate": "2009-11-15T13:12:00",
    "message": "Trying out Elasticsearch, so far so good?"
}'


curl -XPUT 'localhost:9200/bar/_settings' -d '
{
    "index" : {
        "slow_logging.source" : true
    }
}'


curl -XPUT 'http://localhost:9200/bar/tweet/3' -d '
{
    "user": "kimchy",
    "postDate": "2009-11-15T13:12:00",
    "message": "Foo bar baz"
}'

[nik9000]

Space after the equals. Its a silly small change but it helps my eyes.

[nik9000]

Do we really need both the getter style and the bare style?

[bookshelfdave]

probably not. I copied the style of the other settings

[nik9000]

Oh thats cool then. Icky but cool.

[nik9000]

I wonder if it'd be possible to make the setting true/false/some_number_of_characters. If the document has more than that many characters then you could just truncate it. You have to figure out truncation in utf-16 unicode characters but I imagine Lucene has a function for that. I like truncating because usually the neat parts of the document are at the front....

On the other hand if this gets any more complicated than it is it'll really need testing. Could you maybe make a constructor for IndexingSlowLog that takes an ESLog as a parameter? Then you could mock it and get the results. Or something like that.

[jpountz]

+1 to making it take an integer instead of a boolean and truncating. Lucene doesn't have utility methods for that, but Java has: we could just read one more UTF16 char if the last one happens to be a high surrogate?

[nik9000]

Ah yes. I remember that. I was thinking about combining codepoints too but
we might be able to ignore those.
On Aug 6, 2015 5:40 AM, "Adrien Grand" notifications@github.com wrote:

+1 to making it take an integer instead of a boolean and truncating.
Lucene doesn't have utility methods for that, but Java has: we could just
read one more UTF16 char if the last one happens to be a high surrogate
http://docs.oracle.com/javase/7/docs/api/java/lang/Character.html#isHighSurrogate%28char%29
?

—
Reply to this email directly or view it on GitHub
#12655 (comment)
.

[nik9000]

I'll take this in a few days if no one else does.

[nik9000]

Obsoleted by #12806.