Add the filters option to int(), float() and to the rest of the VG functions

[MissaouiChedy]

Pain point

When using the existing data generators (functions under the Pollution.VG module such as int, float, etc), it is possible to specify options such as min, max and must_have values. However, it is not possible to filter out generated values.

Proposal

Add the possibility to use the existing filters option that is not taken into account when passed to the Pollution.VG.int function, among others.

Here is an example usage:

Pollution.VG.int(filters: %{ odd: fn n -> rem(n, 2) == 0 end})

The previous filter specification causes the generator to discard odd integers.

It would be nice if we could use the filter option with all the VG functions.

Details

In my understanding, functions under the Pollution.VG module generates a State struct that acts as a specification that Pollution.Generator uses to generate random values.

Looking inside the following summarized State returned by the following call Pollution.VG.int:

%Pollution.State{
       ......,
       filters: %{}, .....}

We can see that some filter facility is already defined, but seems to be disabled.

In the lib/pollution/generator.ex file, we can see under the as_stream function that filters are fetched and applied to the resulting Stream.

However in the Pollution.Generator.Int.create function, the filters: option is ignored when creating the State struct.

Questions

• Why is the filters: ignored?
• Is there any non obvious concerns regarding filters: that should be addressed before implementing them?

Use case

I came across the need to generate positive integers that are powers of two (1, 2, 4, 8, 16, 32 ...) in my property based tests, the filters: option can help in achieving this behavior.

Kind Regards.