The right precisionStep to use for the trie fields was brought up in SOLR-1483 Here are a few quick size tests to try and see what the impact is.
I tested adding 100,000 random integers (index_fullrange tests), and then 100,000 random integers between 0 and 1000 (index_1000 tests) at various precision steps. I only listed the relevant index parts - the frq (contains the documents that match the term) and the tis (the term infos / list of terms). The tii (term index) is just 128tth of the tis so I don't list it, but it's important. yo...@wolverine /cygdrive/c/code/solr/example/solr/data $ ls -l index_*/*.frq -rwxr-xr-x 1 yonik None 213452 Oct 7 18:57 index_1000_ps0/_0.frq -rwxr-xr-x 1 yonik None 1084313 Oct 7 19:00 index_1000_ps4/_0.frq -rwxr-xr-x 1 yonik None 828082 Oct 7 19:02 index_1000_ps6/_0.frq -rwxr-xr-x 1 yonik None 578506 Oct 7 19:04 index_1000_ps8/_0.frq -rwxr-xr-x 1 yonik None 283486 Oct 7 18:35 index_fullrange_ps0/_0.frq -rwxr-xr-x 1 yonik None 1932237 Oct 7 19:11 index_fullrange_ps4/_2.frq -rwxr-xr-x 1 yonik None 1387882 Oct 7 18:44 index_fullrange_ps6/_0.frq -rwxr-xr-x 1 yonik None 1018848 Oct 7 18:40 index_fullrange_ps8/_0.frq yo...@wolverine /cygdrive/c/code/solr/example/solr/data $ ls -l index_*/*.tis -rwxr-xr-x 1 yonik None 10046 Oct 7 18:57 index_1000_ps0/_0.tis -rwxr-xr-x 1 yonik None 10883 Oct 7 19:00 index_1000_ps4/_0.tis -rwxr-xr-x 1 yonik None 10289 Oct 7 19:02 index_1000_ps6/_0.tis -rwxr-xr-x 1 yonik None 10138 Oct 7 19:04 index_1000_ps8/_0.tis -rwxr-xr-x 1 yonik None 884960 Oct 7 18:35 index_fullrange_ps0/_0.tis -rwxr-xr-x 1 yonik None 3537872 Oct 7 19:11 index_fullrange_ps4/_2.tis -rwxr-xr-x 1 yonik None 2472529 Oct 7 18:44 index_fullrange_ps6/_0.tis -rwxr-xr-x 1 yonik None 2015448 Oct 7 18:40 index_fullrange_ps8/_0.tis Thoughts? -Yonik http://www.lucidimagination.com