Throw a clear error for degenerate-length data vectors (fixes #121)

[JohnCobbler]

quantile! throws ArgumentError("empty data vector") when length is 0. Fixes #121.

typemin(Int):typemax(Int) reports length 0. isempty is false, so the old emptiness check let it through to @assert n > 0. The guard is length(v) == 0.

r = typemin(Int):typemax(Int)
quantile!([0], r, [1])        # ArgumentError
quantile(r, 0.5; sorted=true) # ArgumentError

quantile(r, 0.5) with sorted=false still errors in Base.copymutable.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.43%. Comparing base (e7bf7a4) to head (3f8b0f0).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #206   +/-   ##
=======================================
  Coverage   96.43%   96.43%           
=======================================
  Files           2        2           
  Lines         449      449           
=======================================
  Hits          433      433           
  Misses         16       16           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[andreasnoack]

I believe that such a vector would take up 148 exabytes, so probably not the right recommendation.

[JohnCobbler]

Agreed. 148 exabytes is not a useful suggestion. The message says the length doesn't fit in Int, and asks for a collection whose length does.

[nalimilan]

Thanks! Though I have the impression that printing a specific message for this particular range is overkill. There are probably many places in Statistics, and other JuliaStats packages where we would have to do the same. Yet it seems very unlikely that people would use such a range in practice. So while being correct is of course a requirement, providing a nice error message doesn't seem super important.

How about just replacing the isempty(v) check with length(v) == 0? Or maybe even better, try to fix this in Base, along the lines of JuliaLang/julia#45389 (comment).

[JohnCobbler]

Fair point. The guard is length(v) == 0 now, and this range gets the same empty-vector error as Int[]. I'll leave the Base length change to JuliaLang/julia#45389.

[nalimilan]

Thanks!