fix flaky CSV test
verified with go test -run=Test_CSV -count=1000 ./
The fix was simply to explicitly specify the order of columns in the
SELECT query.
fix a bug which could cause an infinite loop on an improperly terminated quoted string
fix behavior with empty keys
fix up some documentation
implement CSV output support
more informative error messages on failing to open input or output
try adding async reading support
This tanks performance.
Revert performance "improvements"
switch to sorted string lists
re-use key and value lists while parsing
eliminate some uses of maps
This seems to have caused a marginal performance improvement, based on
profiling.
stop allocating arg list in the inner loop