Road to 1.0

[timholy]

I'd like to move the whole JuliaImages stack to version numbers >= 1.0 (see https://github.com/JuliaImages/Images.jl/issues/825). Let's use this issue to collect plans for breaking changes in the near future.

Breaking changes include things like renamings, returning a different value than we used to (e.g., #129 might qualify, though rounding errors are a big ambiguous), or removing externally-visible functionality. Bug fixes, new features, performance improvements, etc., do not count as breaking, though of course they are very welcome.

[kimikage]

Although the specific implementation method needs to be discussed separately, "Overflow checked arithmetics #41" needs to be settled.

I am a speed demon as you know.:smiling_imp: I don't think it is good to throw exceptions during calculation. Perhaps what we need is something like Base.Checked. It would be nice to have a macro similar to @fastmath, which allows the user to specify the types of methods, i.e. *op* or checked_*op*. Of course, I want them but I do never want to write them!

[timholy]

Decide #127

[kimikage]

rand should be return Array instead of ReinterpretArray for the moment(cf. #125).

[timholy]

I like the idea of it returning an Array. Though I also want to fix ReinterpreteArray. I'm thinking of writing a special iterator for it, I think that would theoretically solve the problem at least for the cases we care about in JuliaImages.

[kimikage]

• Drop support for Fixed{Int8,8} and so on (cf. issue #155, PR #159)

[kimikage]

My draft plan

cf. https://discourse.julialang.org/t/rfc-what-should-the-arithmetic-within-the-fixedpointnumbers-be/46697/43 https://discourse.julialang.org/t/rfc-what-should-the-arithmetic-within-the-fixedpointnumbers-be/46697/47

milestone v0.9.0

• [x] Change the default * for Normedｓ to wrapping_mul (PR #236)
• [ ] Move arithmetic definitions/implementations into a submodule FixedPointNumbers.FixedPointArithmetic (issue #239)
  • Export FixedPointArithmetic and Base operators
  • Make wrapping_* and saturating_* public via CheckedArithmeticCore
• [ ] Reserve the aliases for unsigned Fixed/signed Normed (issue #228)
• [ ] Update README
  • Announce checked_*, wrapping_* and saturating_* support
  • Preannounce unsigned Fixed/signed Normed(issue #199)

milestone v0.9.x

• Supports the rest of the issue #185
• Improve Fixed-->Fixed/Normed-->Normed conversions (issue #140)
• Add partial support for 3-arg {wrapping|saturating}_{add|sub|mul}
• Change the type definitions to Fixed{T<:Integer, f}/Normed{T<:Integer, f} internally
  • Open access to Fixed{<:Unsigned}/Normed{<:Signed} only via their aliases
  • Block the use of Fixed{<:Unsigned}/Normed{<:Signed} via constructors and promotions
• Add internal support for Fixed{<:Unsigned}/Normed{<:Signed}

milestone v0.10.0

• ~Make 2-/3-arg wrapping_*/saturating_* public via CheckedArithmeticCore~ -> v0.9.0
• Support Fixed{<:Unsigned}/Normed{<:Signed} officially
  • Force unsupported downstream packages to declare Fixed{<:Signed}/Normed{<:Unsigned}
• Drop support for Fixed{Int8,8} and so on
  • cf. issue #155, PR #159

milestone v0.11.0

• Determine the interface for using multiple types of arithmetic

milestone v1.0.0

• Remove deprecated stuff

[kimikage]

My personal roadmap has not changed for three years. However, because so much elapsed time has passed, I have withdrawn the roadmap. In other words, I have removed some v0.9 milestone flags.

Therefore, we may be able to release v0.9 without the saturating_* and so on. (I proposed them as a solution to a problem, but the parties to the problem have little interest in the development.)

Meanwhile, unsigned Fixed is something I would like to see its definition by v0.9. (Its functionality can be implemented in downstream packages.)