This is a big one. I am porting HTTP server to use CAIRO storage system. This turned out to be a complete rewrite. I just started testing ported code to realise the number of gaps, bugs and inconsistent behavior in code that imports text files into database. I am aiming to get test coverage to 100% and test all use cases I can think of right now. I am not far off the target but there are few more things still to be completed. Here is summary of the change:

- ported text file parser and import mechanism to CAIRO
- added a tonne of tests to get code coverage to 87% for now
- fixed a number of bugs with encoding two 32-bit integers as single 64-bit one
- consolidated table creation code (there were 4 copies, now there is one working consistently)
- copied HTTP server into "tuck" package, which is where new server will be evolving before it moved elsewhere

GRIFFIN: symbol type cast for create table statement supports caching option and validates symbol capacity value

CAIRO: Null for symbol value now corresponds to null for int. This removes unnecessary complexity around implementing getInt()

CAIRO: bug fixes

GRIFFIN: initial limit implementation, e.g. select * from tab limit 10. It supports last N rows as well as first N rows. More testing to follow.

LINUX: C code refactoring to allow for different implementations of interrupting of blocking accept()

GRIFFIN: bugfixed NPE when select had fewer columns than target table needed and columns are ambiguous.

GRIFFIN: join implementations propagate timestamp column from master if it exists. ASOF join validates that both master and slave cursors have timestamp defined. Validation errors are reported using position of "join" keyword in query text and referring to left or right side of this join.