A Python library for calculations over large arrays of numbers: what ordinary code would take minutes to compute, numpy does in a fraction of a second.
In ToshaMusic numpy breaks a fragment of a track down by frequency (FFT) and finds the spectral cut-off. That is how the bot catches a FLAC rebuilt from an MP3 and a CD master stretched to 24-bit / 192 kHz.
When data must not just be stored but crunched heavily and fast: signal analysis, statistics over large arrays, file quality checks.
In ToshaMusic the bot checks that a downloaded FLAC really is Hi-Res. A fragment from the middle of the track is decoded to uncompressed audio, and numpy breaks it down by frequency (an averaged FFT) and finds the frequency above which the spectrum is empty. A cut-off in the 14 to 20.6 kHz range gives away a file rebuilt from an MP3, and nothing above 24 kHz on a "96/192 kHz" release reveals a stretched CD master.
For ordinary record-keeping, orders and reports, a database and its queries are enough: numpy is for when the calculations are many and heavy.
Terms explained โ now let's get to work: tell us about the task and we'll turn it into a clear work plan.