Details
-
Bug
-
Resolution: Fixed
-
Critical
-
2.0.1
-
Security Level: Public
-
None
-
Tested on OS X but easy to reproduce on any machine
1 single node cluster
-
MacOSX 64-bit
-
n/a
Description
I have created a very simple view that I want to use to do some queries "document ID". This is equivalent to our all_docs view
function (doc, meta)
{ emit(meta.id); }The issue is the fact that even if my view does not do anything with the "doc", the size of the doc itself impact a lot the indexing time. I have only tested with "initial indexing", that could become very quickly not manageable.
Here some basic numbers:
ms |nb of docs | doc size |
28 | 10 |1.0 kB |
43 | 100 |1.0 kB |
490 | 10 |512.0 kB |
4674 | 100 |512.0 kB |
989 | 10 | 1.0 MB |
9642 | 100 | 1.0 MB |
1972 | 10 | 2.1 MB |
19382 | 100 | 2.1 MB |
9750 | 10 |10.5 MB |
97325 | 100 |10.5 MB |
PS: I know/understand that we are doing incremental map reduce, and it will be not that big once the index is created. It would be great if we can avoid sending the "doc" when it is not necessary.