[False Positive] Json file detected as SQL

[emanuelb]

Running guesslang data-2.0.json return Programming language: SQL
while running guesslang --probabilities data-2.0.json return: (the other langs have 0%)

Language name       Probability
 SQL                  98.52%
 JavaScript            1.31%
 PHP                   0.16%
 Python                0.01%

The file is:
https://sources.debian.org/data/main/n/node-caniuse-db/1.0.30001048-1/fulldata-json/data-2.0.json

while running guesslang on file data-1.0.json from:
https://sources.debian.org/data/main/n/node-caniuse-db/1.0.30001048-1/fulldata-json/data-1.0.json
works somehow correctly by returning Programming language: JavaScript & JavaScript 100.00% (as JSON not supported yet, but it's valid JavaScript)

[yoeo]

Hi @emanuelb

Indeed, JSON is not supported yet. The current model was not trained with JSON files see:

guesslang/guesslang/data/languages.json

Line 13 in 5537217

"JavaScript": ["js", "es6"],

In consequence, even if Guesslang may sometimes correctly classify JSON as JavaScript, classification errors are highly expected here.

A proper support for JSON files is planned #19, I just need to find enough JSON example files to feed the model.

[yoeo]

Hello @emanuelb ,

Guesslang properly supports JSON now as well as 23 other languages that where not supported before #33 🙂