12bitfloat 5c50a45101 fix: comment out slow python3 benchmark section in bench.py
The Python benchmark using isspam.py was removed from the benchmark
run because its execution time was excessively slow compared to the
C++ and Rust implementations, skewing overall timing results.
2025-10-06 22:56:44 +00:00

isspam

Fast as light evaluator for text files to summarize specific details about the text files.

This repository contains multiple versions of the same(-ish) algorithm.

Versions

Building

Build all versions to the repo root:

make build_all

Build isspam (C) with memory check (requires valgrind to be installed):

make valgrind

Benchmarking

After all binaries have been build to the repo root, you can benchmark them like this:

make benchmark

or without extracting books again:

make benchmark_only

Running

Using files as parameter

./(r)isspam ./spam/*.txt
./(r)isspam ./not_spam/*.txt

Using stdin

Useful for automation. Works only on the isspam version.

cat ./spam/example_spam1.txt | ./isspam

Example output

Output example made by isspam.

File: ./spam/example_spam3.txt
Capitalized words: 39
Sentences: 20
Words: 420
Numbers: 1
Forbidden words: 15
<0:recovery>
<1:techie>
<2:https>
<3:digital>
<4:hack>
<5://>
<6:com>
<7:@>
<8:crypto>
<9:bitcoin>
<10:whatsapp>
<11:cryptocurrency>
<12:stolen>
<13:contact>
<14:understanding>
Word count per sentence: 21
Memory usage: 1 MB, 6.460 (re)allocated, 4.222 unqiue free'd, 0 in use.

Valgrind status

Valgrind output for isspam version.

Rust variant thinks it's too cool for memory checks afterwards.

Date: 2024-11-30

==58062== 
==58062== HEAP SUMMARY:
==58062==     in use at exit: 0 bytes in 0 blocks
==58062==   total heap usage: 6,490 allocs, 6,490 frees, 2,343,156 bytes allocated
==58062== 
==58062== All heap blocks were freed -- no leaks are possible
==58062== 
==58062== For lists of detected and suppressed errors, rerun with: -s
==58062== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)
S
Description
For a site I was doing spam analysis, how to recognize it. I do this by checking how it is written. How much numbers are used. How many capitals. That's the first indication. I made it originally myself but a few people liked the project and wrote their versions in their preferred language. Very happy about that. We are bench marking it as challenge, the benchmark is doing analysis for 900 books in txt format. The fastest application wins. By now, every language has as wel won as lost. We keep iterating and making them faster. Thanks 12bitloat, BordedDev en JestDotty for contribution of their source.
Readme
486 MiB
Languages
Rust 73%
C++ 12.9%
C 6.8%
Swift 3.4%
Python 2.2%
Other 1.6%