Here is a screenshot of a query and the results I'm seeing:
It is exceptionally frustrating because it worked for one of the
values once, but not the other. I restarted ES to see if it was a
glitch, then it was not working for either. No config changes,
nothing other than a shutdown and restart of ES.
I created the following mapping template and validated that every
wildcarded entry properly shows as type: nested and include_in_root:
Daniel, try and use the full nested path for the fields you specify in the
facets. Note, if you specify include_in_parent, those facets will require
considerable amount of extra memory, so, try and see if you really need
it...
It is exceptionally frustrating because it worked for one of the
values once, but not the other. I restarted ES to see if it was a
glitch, then it was not working for either. No config changes,
nothing other than a shutdown and restart of ES.
I created the following mapping template and validated that every
wildcarded entry properly shows as type: nested and include_in_root:
Yes. I was not specifying the full path for key_field and
value_field. That made it work.
Now that I have at least one working case, I'll go back and try again
without include_in_parent. that was just a wild stab at getting it to
work.
Question regarding performance (assuming that they are not included
in parent/root):
If I have hundreds of thousands or even millions of root documents,
and dozens of this histogram objects, each with tens of nested
"values" objects, I will end up with a technical document count in the
billions because nested objects are indexed as extra documents. Will
I suffer in terms of scale in this case or are these nested object
documents efficient enough that I don't have to worry about having
billions when I only have hundreds of thousands of parent docs?
Daniel, try and use the full nested path for the fields you specify in the
facets. Note, if you specify include_in_parent, those facets will require
considerable amount of extra memory, so, try and see if you really need
it...
It is exceptionally frustrating because it worked for one of the
values once, but not the other. I restarted ES to see if it was a
glitch, then it was not working for either. No config changes,
nothing other than a shutdown and restart of ES.
I created the following mapping template and validated that every
wildcarded entry properly shows as type: nested and include_in_root:
Well, you will need to test it out (and scale if perf is not good
enough). In general, those facet values are loaded to memory, and
computation is pretty fast, so, we make a lot of effort to make it as fast
as possible...
Yes. I was not specifying the full path for key_field and
value_field. That made it work.
Now that I have at least one working case, I'll go back and try again
without include_in_parent. that was just a wild stab at getting it to
work.
Question regarding performance (assuming that they are not included
in parent/root):
If I have hundreds of thousands or even millions of root documents,
and dozens of this histogram objects, each with tens of nested
"values" objects, I will end up with a technical document count in the
billions because nested objects are indexed as extra documents. Will
I suffer in terms of scale in this case or are these nested object
documents efficient enough that I don't have to worry about having
billions when I only have hundreds of thousands of parent docs?
Daniel, try and use the full nested path for the fields you specify in
the
facets. Note, if you specify include_in_parent, those facets will require
considerable amount of extra memory, so, try and see if you really need
it...
It is exceptionally frustrating because it worked for one of the
values once, but not the other. I restarted ES to see if it was a
glitch, then it was not working for either. No config changes,
nothing other than a shutdown and restart of ES.
I created the following mapping template and validated that every
wildcarded entry properly shows as type: nested and include_in_root:
Daniel,
I am facing a similar situation. I am wondering if you learned something
from your experience. If you can share any interesting bits found, then it
would really help. thanks!
Vinay
On Tuesday, August 16, 2011 7:46:22 PM UTC-7, Daniel E wrote:
Will definitely be testing it. Just a bit scary to go from 1 million docs
to 5 billion just by setting type:nested.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.
(and scale if perf is not good