Introduction

StreamPref is a data stream management system (DSMS) prototype with support to temporal conditional preferences queries StreamPref Project.

The execution of queries by the system is controlled by an environment file. On following, we have an example of a environment file:

###############################################################################
# Registration of tables and streams
#
# You have to inform the table/stream name, attributes, attributes types and
# the input file (for data reading). 

# Stream example
REGISTER STREAM s (a INTEGER, b STRING, c FLOAT)
INPUT 'in/s.csv';

# Stream table example
REGISTER TABLE t (a INTEGER, b STRING, c FLOAT)
INPUT 'in/t.csv';
###############################################################################
# Registration of queries
#
# You have to inform the query name, the input file with the query code.
# Optionally, you can inform an output file to print the query result.

REGISTER QUERY q
INPUT 'queries/q.query'
OUTPUT 'out/q.out';

# You also can use the keyword CHANGE to print the insertions and deletions
# instead of the query result

REGISTER QUERY q2
INPUT 'queries/q2.query' 
OUTPUT CHANGES 'out/q2.out';

Every input file must have the column _fl with the instant (or timestamp) of the tuple and the remaining columns. The name of the columns must be identical to attributes name declared in the registration of the stream/table. Example of an input stream file:

_ts,  a, b,  c
  0,  1, AA, 0.5
  0,  2, BB, 0.6
  1,  1, AA, 0.5
  2,  3, CC, 1.6
  2,  1, AA, 0.7
  2,  2, BB, 0.8
  3,  1, AA, 0.9
  4,  3, CC, 1.0

The input table files must have the additional attribute _fl. This attributes indicates the insertion (+) or deletion of a tuple (-). We can delete only those tuples inserted in previous instant. If you try to delete an non-existing tuple, the system raises an exception. Example of an input table file:

_ts, _fl, a, b,  c
  0, +, 1, AA, 0.5
  0, +, 2, BB, 0.6
  1, -, 1, AA, 0.5
  2, +, 3, CC, 1.6
  2, +, 1, AA, 0.7
  2, -, 2, BB, 0.6
  3, +, 1, AA, 0.9
  4, +, 3, CC, 1.0

The query file content must follow the StreamPref grammar (see file grammar/streampref.ebnf). Example of an query file with temporal conditional preferences:

SELECT SUBSEQUENCE END POSITION
FROM SUBSEQUENCE CONSECUTIVE TUPLES
FROM SEQUENCE IDENTIFIED BY pid [RANGE 20 SECOND]
FROM moves AS m
WHERE MINIMUM LENGTH IS 2
AND MAXIMUM LENGTH IS 4
ACCORDING TO TEMPORAL PREFERENCES
IF FIRST THEN move = 'rec' BETTER move = 'lbal'
AND
IF PREVIOUS (move = 'cond') THEN (move = 'drib') BETTER (move = 'pass')[place]
AND
(move = 'pass') BETTER (move = 'bpas')[place]
AND
IF ALL PREVIOUS (place = 'oi') THEN (place = 'oi') BETTER (place = 'mf');

Please see the directory examples for more examples.

Command Line

streampref.py [-h] -e ENV [-m MAX] [-l LOG] [-D] [-o OUTCOMP]
              [-d DETAILS] [-r DELIMITER] [-p PREF_ALG] [-t TPREF_ALG]
              [-s SUBSEQ_ALG]
  -h, --help                              Show the help message and exit
  -e ENV, --env ENV                       Environment file
  -m MAX, --max MAX                       Maximum timestamp to run
  -l LOG, --logfile LOG                   Log file
  -D, --debug                             Debug execution
  -o OUTCOMP, --outcomparisons OUTCOMP    Output comparisons statistics
  -d DETAILS, --details DETAILS           Save execution details to file
  -r DELIMITER, --delimiter DELIMITER     File content delimiter
  -p PREF_ALG, --pref PREF_ALG            Preference algorithm
  -t TPREF_ALG, --tpref TPREF_ALG         Temporal preference algorithm
  -s SUBSEQ_ALG, --subseq SUBSEQ_ALG      Subsequence algorithm

Algorithms

When the queries have the operators BEST, TOPK, CONSEQ, ENDSEQ, BESTSEQ and TOPKSEQ, the user can choose an algorithm to evaluate the operators. The evaluations of the operators BEST and TOPK can be performed by the following algorithms:

depth_search: Depth search first comparison algorithm (non incremental);
partition: Preference partition algorithm (non incremental);
inc_ancestors: Ancestors list algorithm (incremental);
inc_graph: Graph algorithm (incremental);
inc_graph_no_transitive: Graph algorithm without transitive comparisons (incremental);
inc_partition: Preference partition algorithm (incremental).

The evaluation of operators CONSEQ and ENDSEQ can be performed by the following algorithms:

naive: Naive algorithm (non incremental);
incremental: Incremental algorithm using deletions and insertions.

The evaluation of the algorithms BESTSEQ and TOPKSEQ can be performed by the following algorithms:

depth_search: Depth search first algorithm (non incremental);
inc_seqtree: Algorithm using the sequences tree index (incremental);
inc_seqtree_pruning: Algorithm using the sequences tree index with pruning strategy (incremental).

Please see the related publications for more information.

StreamPref DSMS Prototype

StreamPref Data Stream Management System (DSMS) prototype.

Table of Contents

Introduction

Command Line

Algorithms