Resql Database System and Flounder Library (Prototypes)
Authors: Henning Funke <henning.funke@cs.tu-dortmund.de>
Jan Mühlig <jan.muehlig@tu-dortmund.de>
Query compilation-based database system with low compilation times.
Research
ReSQL and Flounder are used as prototype systems for database
research. This research includes the following publications:
Funke, Henning, Jan Mühlig, and Jens Teubner. "Low-latency query
compilation." The VLDB Journal (2022).
Funke, Henning, and Jens Teubner. "Low-latency compilation of SQL
queries to machine code." Proceedings of the VLDB Endowment (2021).
Funke, Henning, Jan Mühlig, and Jens Teubner. "Efficient
generation of machine code for query compilers." DaMoN (2020).
Build
First make sure that the libraries and tools are available.
Then use 'make' in this folder. Developed and tested under
Ubuntu 18.04. For debug builds use 'make DEBUG=1'. To run
tests use 'make run-tests'.
Libraries
flex needs to be installed, e.g. via apt
libreadline needs to be installed, e.g. via apt
asmjit source included, build with cmake in lib/asmjit
lemon source included, no build necessary
cereal source included, no build necessary
cxxopts source included, no build necessary
Other tools
nasm assembler
clang c++ compiler
Run
ReSQL runs either in interactive or client/server mode. Use './resql -h'
for instructions. If the file startup.sql exists in this folder, ReSQL
will execute it at the start of the server (or standalone). A startup
file with the following commands, for example, will load a sample
database (10 MB).
showplan=true;
exec tpch/create.sql;
exec tpch/load_sf001.sql;
Commands
showfln=true show Flounder IR
showasm=true show machine assembly
showperf=true show performance measurements
tables show tables from current database
emitmc=true assemble via asmjit
emitmc=false assemble via nasm
threads=4 use 4 threads for execution
tofile=true write query results to file
qres.tbl (server)
exec file.sql execute file with SQL statements
separated by ';'
select * execute SQL
q or exit quit
ReSQL understands the following commands (for example)
create table name ( name1 type1, name2 type2 )
bulk insert name from "path/foo.tbl" with ( fieldterminator="|" )
select c,avg(d*a) from foo,bar where a=d group by c order by c
Known Issues
- Memory for Expr ist not freed.