Add single argument version of isapprox

[asinghvi17]

Adds a single argument version of isapprox or ≈, in the same vein as the single argument version of ==.

[asinghvi17]

@fredrikekre, why are you confused?

[StefanKarpinski]

Looks good to me. Could you add some tests and a NEWS entry?

[ararslan]

The Documenter ref there is self-referential. Probably doesn't matter a whole lot, but you could have it point to the other method specifically.

[asinghvi17]

How do you do that - with the method signature? So something like [\isapprox(1, 2)`]`?

[ararslan]

I'm not 100% on the syntax, but I think it'd be something like

[`isapprox`](@ref Base.isapprox(::Any, ::Any))

You'd probably want to build the documentation locally to verify.

[JeffBezanson]

Or just say "The keyword arguments supported are the same as those in 2-argument isapprox."

[asinghvi17]

Thanks for the review, @ararslan!

[nickrobinson251]

I'm surprised not to find this pattern - like all(≈(x), xs) - in the base test code more. A quick search suggested it is only used once

julia/test/error.jl

Line 8 in 90d481e

@test all(x->x ≈ 10.0, ratio(collect(ExponentialBackOff(n=10, max_delay=Inf, factor=10, jitter=0.0))))

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@7e7a30f). Click here to learn what that means.
The diff coverage is 84.09%.

@@            Coverage Diff            @@
##             master   #32305   +/-   ##
=========================================
  Coverage          ?   75.74%           
=========================================
  Files             ?      380           
  Lines             ?    66442           
  Branches          ?        0           
=========================================
  Hits              ?    50326           
  Misses            ?    16116           
  Partials          ?        0

Impacted Files	Coverage Δ
base/util.jl	48.72% <ø> (ø)
base/coreio.jl	50% <ø> (ø)
base/operators.jl	87.7% <ø> (ø)
base/initdefs.jl	54.95% <ø> (ø)
base/deprecated.jl	9.3% <ø> (ø)
base/refpointer.jl	65.9% <ø> (ø)
base/special/log.jl	94.95% <ø> (ø)
base/sort.jl	94.25% <ø> (ø)
base/logging.jl	87.66% <ø> (ø)
base/grisu/grisu.jl	16.82% <ø> (ø)
... and 81 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7e7a30f...52f93fa. Read the comment docs.

[nickrobinson251]

thanks! I would find this useful :)

Is there any reason not to use the Fix2 machinery, same as with isequal?

[StefanKarpinski]

I'm surprised not to find this pattern - like all(≈(x), xs) - in the base test code more.

If the intention is approximate numerical equality, then x ≈ xs is more appropriate as well as more convenient: it checks that norm(x .- xs) is less than an appropriate error bound.

[JeffBezanson]

Ready to merge.

[nickrobinson251]

If the intention is approximate numerical equality, then x ≈ xs is more appropriate as well as more convenient: it checks that norm(x .- xs) is less than an appropriate error bound.

Thanks. Good point :) Somewhat off topic here (sorry), but I was thinking of the case where x is scalar, xs is Vector (in which case x ≈ xs is a MethodError, and all(≈(x), xs) may be what one wants).