Trinity-Edge

Exploring faster deployments of Trinity for the paper "[Trinity: An Extensible Synthesis Framework for Data Science](https://dl.acm.org/doi/10.14778/3352063.3352098)".

This is a customized/optimized version of the original Trinity (https://github.com/fredfeng/Trinity), with some additional features that support efficiency and research purpose.

Prerequisites

• Python 3.8.8 (tested)
• xxhash 2.0.2 (tested)
• sexpdata 0.0.3 (tested)

Quick Commands

# testing a benchmark with its target skeleton only
python ./test_morpheus_benchmark.py --benchmark 5 --dsl test_min --skeletons test_min

# testing a benchmark with shared skeleton list
python ./test_morpheus_benchmark.py --benchmark 5 --dsl test_min --skeletons ngram3_nojoin

Usage

single benchmark

usage: test_morpheus_benchmark.py [-h] [-b BENCHMARK] [-d {test_min}] [-s {test_min,test_ks0,ngram3_nojoin}] [--record]

optional arguments:
  -h, --help            show this help message and exit
  -b BENCHMARK, --benchmark BENCHMARK
                        morpheus benchmark id, default: 5
  -d {test_min}, --dsl {test_min}
                        DSL definition to use, default: test_min
  -s {test_min,test_ks0,ngram3_nojoin}, --skeletons {test_min,test_ks0,ngram3_nojoin}
                        skeleton list to use, default: test_min
  --record              whether to enable recorder or not, default: False

Design Notes

• For a list of columns in the parameter, sort it ASC, and apply the semantics (negative: exclude, positive: include) according to the order.

TODOs / Known Issues

• spread/group component are much slower than others.
• Test all morpheus benchmarks to make sure they can actually be solved.
• Sketch level CDCL is not yet ready.
• Automatically synthesize the abstract interpreter based on the provided language specs.
• To speed up parameter enumeration and improve efficacy, consider letting newly generated columns stay in front (not at the end).

Citation

If you find our work and this tool useful in your research, please consider citing:

@article{vldb19_trinity,
  author    = {Ruben Martins and
               Jia Chen and
               Yanju Chen and
               Yu Feng and
               Isil Dillig},
  title     = {Trinity: An Extensible Synthesis Framework for Data Science},
  journal   = {{PVLDB}},
  volume    = {12},
  number    = {12},
  pages     = {1914--1917},
  year      = {2019},
  url       = {http://www.vldb.org/pvldb/vol12/p1914-martins.pdf},
  timestamp = {Fri, 30 Aug 2019 13:15:08 +0200},
  biburl    = {https://dblp.org/rec/bib/journals/pvldb/MartinsCCFD19},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}