ta-lib-python VS bcolz

One such excellent Python library is TA-Lib. (https://github.com/TA-Lib/ta-lib-python)

Have you looked TA-Lib?

TA-Lib

Dev here. I've wanted to get into doing this for as long as I can remember. Back in 2018 I got a half start at it, pulling in Binance candle data for all intervals with python and then running them all through every TA function of ta-lib to generate corresponding indicators. I intended to take it that next step further into picking up ML, and wanted to write a script that would pair every permutation of indicator looking for combinations that improve certainty for price direction changes (the certainty of BOLL + RSI was going to be used as my baseline for a known semi-reliable combo). I would really love to circle back someday to finishing that up, and then start sucking in candle data from as many markets as I could afford the AWS services for lol. That's my dream dashboard.

Not sure what they are using specifically, but ta-lib has candle pattern matching.

Just use TA-Lib https://mrjbq7.github.io/ta-lib/

Doesn't seem like an official mirror but if you're curious what the code looks like, including the python/C bindings, here it is: https://github.com/mrjbq7/ta-lib

I use talib https://mrjbq7.github.io/ta-lib/ and it accurately mirrors binance

This from a closed issue.

What you need for your use case is a column-oriented store. I recommend explore bcolz or apache arrow for a column file-based systems. These are very fast, support memory mapping, uses compression and SSD speed (and even CPU architecture, in case of arrow) optimally almost out of the box, and has good interfaces to Numpy and Pandas (in case you are using Python for final data consumption and analysis). The columnar structure makes it easy to add or delete a column easily (or even dynamically). If you need a more scalable (albeit at the cost of speed) solution, you can devise a schema over a regular columnar db or an nosql db - see arctic from Man group for an example.