Stream-processing-benchmark

Finding the best way to process stream of text data for each programming language.

This is a benchmark for testing different coding patterns for different languages. The test is about text parsing and processing speed.

To run the benchmark, simply go to the bench directory and the type:

bash benchmark.sh

It should run six programs:
- (optional) datagenerator will generate a data set file (which, I hope will be the same for everybody). The dataset stays between run, so this program is called once.
- Three PARSERs (one of them producing an incorrect result)
- Two PROCESSORs

You can compare with my results:
cat benchmark_report_fade_laptop.txt
cat benchmark_report_fade_desktop.txt

TRY YOUR OWN!

THE DATASET:
The dataset is a four column set as follow:
integer[tab]string1[tab]string2[tab]float
It may contains commentary lines (beginning with #) and malformed lines (wrong number of fields or fields containing wrong type). For simplicity, float is only integer with an optional [.integer] part.

THE RULES:
There is two groups of programs:

PARSER
A PARSER takes the dataset as input and produce the output described bellow and some statistics:

F1[tab]F2[tab]F3[tab]F4 -> F1,F4,F3,F2

Every non conforming (and comment) line is discarded. The end of the ouput is a stat line like this one:
parsed line: X, commentary line: Y, unknown line: Z. You shall figure out X,Y, and Z :)

In doubt, the PERL SIMPLE PARSER is the reference for the parser category. Except for # lines, the output of the PARSER must be the same than PERL SIMPLE PARSER.

EXAMPLE :
$ cat common/tinydataset.txt
# This file is for demonstration only
1	This	IsAnExample	1
2	This	IsTooAnExample	2.4
2	This	IsTooAnExample	2.7
1	That	IsAMalformedline
3	That	IsAMalformedline	Too
$ cat common/tinydataset.txt | perl perl5/simpleparser.pl 
1,1,IsAnExample,This
2,2.4,IsTooAnExample,This
2,2.7,IsTooAnExample,This
#PERL version 5.14.2
parsed line: 6, commentary line: 1, unknown line: 2.

Notice that the PERL line is prefixed by #: it will be ignored by the conformity check. The last line, with statistics is part of the output.

PROCSSOR

A PROCESSOR takes the same dataset but and do the same thing than the parser, rounding the value instead of reproducing it, and then produce a list of aggregation data.

Internally you have a first key (second field of input), a second one (third field of input), and a value (fourth one)

The aggregation is a sorted list of sorted key with a sum of all value. The sum is made in float but display as an int (not rounded: truncated).

input:
A,B,C,D
E,B,F,G
H,B,C,I
output:
B: (C:int(D+I)) (F:G)

EXAMPLE:

$ cat common/tinydataset.txt | perl perl5/simpleprocessor.pl 2> /dev/null
#PERL REFERENCE PROCESSOR
#TRANSFORMED INPUT
1,1,IsAnExample,This
2,2,IsTooAnExample,This
2,3,IsTooAnExample,This
#DATADUMP
This: (IsAnExample:1) (IsTooAnExample:5)
#REPORT
#PERL REFERENCE PROCESSOR
#perl version: 5.14.2
parsed line: 6, commentary line: 1, unknown line: 2, keystat: 2.