Port library from `BigDecimal` to `Double`, and add QoL improvements

ggalmazor commented 9 months ago

Fixes #23

Problem to be solved:

BigDecimal is not an optimal solution for large data series, causing excessive memory consumption. Better alternatives can be used without losing precision.
The library is too opinionated regarding the input types that must be provided. Users need to either use or extend Point, which is a concrete class, and they are bound to its internal types.
The current implementation has some redundant bits that can be trimmed off

In this PR:

Port Point from BigDecimal to Double
Port the Point factory to accept Number objects for maximum compatibility
Move the Point class to DoublePoint and extract its interface into Point
Move point operations on DoublePoint objects (add, subtract, half) to Point and make static
Migrate tests to a reusable custom matcher that compares actual x and y values instead of relying on equality.
Remove the custom equals and hashCode everywhere. This was only needed for the tests.
Random minor cleanup

capsuleman commented 9 months ago

It looks good for me to merge your PR, is it ok for you?

ggalmazor commented 9 months ago

Sorry, @capsuleman , I've been super busy with work this week. Let's merge it a roll out a 0.1.0 version

ggalmazor commented 9 months ago

Alright, it's done!

capsuleman commented 9 months ago

No problem! Thank you again for your help, you rock! ⭐

ggalmazor / lt_downsampling_java8