InFine

Coverage

Scripts used to compute coverage are provided in ./coverage/ directory. To compute the coverage, variables SG_FILES and SPJ_FILES in Coverage_script.sh must point to the directory containing base files and query result files, respectively.

Datasets

Due to space limitations, all open datasets (PTC,TPC-H and PTE) used in the experiments are available on the following link: Datasets used in InFine experiments

MIMIC-3 can be obtained on MIMIC-3

Reproducibility and repeatability

Running InFine

InFine runs on Windows and Linux OS. The Linux version of infine can be found in ./exec/linux. The Windows version of infine with the necessary libraries (g++ compiler mingw32) can be found in ./exec/windows.

The file allResultsInFine.txt in ./exec/ contains our statistical results.

Remark: Due to the double compatibility of OS, it is possible that some text files have to be re-encoding (unix2dos or dos2unix).

Run experiments

We provide all files we used in our experiments for PTE (works_PTE.txt), PTC (works_PTC.txt), and TPC-H (works_TPCH.txt) databases. Each works file is located in the corresponding datasets folder (for example, works_PTE.txt in datasets PTE folder, cf. Datasets used in InFine experiments).

Make sure that the path of file in the works file is valid.

Use InFine

The command line to launch InFine prototype consists of either 1 parameter (name of file containing a jobs list) or 9 parameters (the description of a single job).

A job is described by DBName, TableORquery, dataInput1.csv, attrJoinId1, fdsInput1.txt, dataInput2.csv, attrJoinId2, fdsInput2.txt, nbRuns where:

DBname: the database which corresponds to the set of data sets (information)
TableORquery: name of join table or a query (information). Due to some special characters, one needs to enclose by "
dataInput1.csv attrJoinId1 fdsInput1.txt: the LEFT data for the join operation following by the id of the join attribute and the valid FDs
dataInput2.csv attrJoinId2 fdsInput2.txt: the RIGHT data for the join operation following by the id of the join attribute and the valid FDs
nbRuns: number of runs with these parameters. It uses for statistics.

For instance: PTE "pte_atm|X|pte_drug" pte_atm.csv 0 pte_atm_nulleq_efd.txt pte_drug.csv 0 pte_drug_nulleq_efd.txt 10 pte_atm.csv and pte_drug.csv coming from the database PTE. The ids 0 and 0 correspond respectively to the id of the join attribute of pte_atm.csv and pte_drug.csv. pte_atm_nulleq_efd.txt and pte_drug_nulleq_efd.txt contain respectively the valid FDs of pte_atm.csv and pte_drug.csv . These settings are executed 10 times.

A jobs file is a text file that contains jobs description. Empty lines or lines starting by '#' are ignore. For example, the following file contains only one job description.

# InFine for PTE

# "pte_atm|X|pte_drug"
PTE "pte_atm|X|pte_drug" pte_atm.csv    0 pte_atm_nulleq_efd.txt    pte_drug.csv   0 pte_drug_nulleq_efd.txt 10

Infine's results are store in two types of files

a file named DBname_results.txt (PTE_results.txt for example) for statistics usage, one line is appended per run
a file named DBname_TableORquery_provFDs.txt (PTE_pte_atm_JOIN_pte_drug_provFDs.txt, "|X|" are replaced by "JOIN" due to '|' character) for the provenance identification of valid FDs (only once for all runs).

Due to possible duplication of attribute names in datasets, infine rename all attributes by adding prefix: L_ and R_ for LEFT and RIGHT datasets respectively. The prefix JA_ identified the join attribute.

LanLi2017 / InFine