niquola
commented
8 years ago Cool. Please provide tests & documentation in readme for this feature and i will merge it.
And may be simplify namings: _total={no,estimated,exact} another option to move this flag in params:
{"resourceType":"Patient", "total": "estimated", "queryString":"name=...."}
usage: no param - calculates "total" basing on standard COUNT(*) function _totalMethod=std - same as above _totalMethod=improved - calculated approximaged total using count_estimate function - which is using EXPLAIN data _totalMethod=no - do not calculate total at all (fastest but no total data, and no links to next, prev, first and last
results for 1M patients: fhirbase=# explain analyze select fhir_search('{"resourceType":"Patient", "queryString":""}');
QUERY PLAN
Result (cost=0.00..0.26 rows=1 width=0) (actual time=480.715..480.715 rows=1 loops=1) Planning time: 0.012 ms Execution time: 480.725 ms (3 rows)
fhirbase=# explain analyze select fhir_search('{"resourceType":"Patient", "queryString":"_totalMethod=no"}');
QUERY PLAN
Result (cost=0.00..0.26 rows=1 width=0) (actual time=3.415..3.415 rows=1 loops=1) Planning time: 0.025 ms Execution time: 3.434 ms
fhirbase=# explain analyze select fhir_search('{"resourceType":"Patient", "queryString":"_totalMethod=improved"}');
QUERY PLAN
Result (cost=0.00..0.26 rows=1 width=0) (actual time=10.147..10.148 rows=1 loops=1) Planning time: 0.012 ms Execution time: 10.157 ms (3 rows)
although select fhir_search('{"resourceType":"Patient", "queryString":"_totalMethod=std"}'); returns: {"resourceType":"Bundle","type":"searchset","total":1000000,"link":[{" ...
and select fhir_search('{"resourceType":"Patient", "queryString":"_totalMethod=improved"}'); returns: {"resourceType":"Bundle","type":"searchset","total":1000074,"link":[{ ....
which gives quite a satysfying results, but error is might be bigger for smaller volumes of data.