Add parsing support for Interval

[omus]

Adds parsing support for Interval{T}. I ended up just using comma as the parsing delimiter here as I expect it's the syntax most people would expect. Because this PR currently only supports the comma delimiter I changed the printing of intervals to also use a comma. Using a comma for the delimiter when printing has a chance of having a user think that a printed closed-interval is actually a Vector (e.g. [1, 2]). We could work around this by leaving printing using .. and just support both delimiters when parsing.

The parsing performance seems pretty good here despite using a regex. I expect we can improve the parsing performance further but I'm happy with this for now:

julia> using Intervals, BenchmarkTools

julia> @btime parse(Interval{Int}, "[1,2)")
  731.831 ns (13 allocations: 656 bytes)
Interval{Int64,Closed,Open}(1, 2)

julia> @btime parse(Int, "1")
  26.491 ns (0 allocations: 0 bytes)
1

Fixes #116.

Update: Both parsing of delimiters .. and , are supported now and printing has been left unchanged.

[omus]

I know that LibPQ rolled its own interval parsing code so I took a look at it to ensure this PR works nicely there. The only issue I see is that LibPQ wants to call it's own parsing function. I'll try to support that.

[codecov]

Codecov Report

Merging #128 into master will increase coverage by 0.80%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #128      +/-   ##
==========================================
+ Coverage   73.92%   74.72%   +0.80%     
==========================================
  Files           9       10       +1     
  Lines         441      455      +14     
==========================================
+ Hits          326      340      +14     
  Misses        115      115              

Impacted Files	Coverage Δ
src/Intervals.jl	80.00% <ø> (ø)
src/parse.jl	100.00% <100.00%> (ø)

---

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 be20134...82d93e1. Read the comment docs.

[mattBrzezinski]

The parsing performance seems pretty good here despite using a regex.

Are regexes in Julia inherently slow?

[mattBrzezinski]

Just needs a version bump, I assume we would want to bring along a few of the other CRs across the line before making a new release though.

[omus]

The parsing performance seems pretty good here despite using a regex.

Are regexes in Julia inherently slow?

The issue isn't that regexes in Julia are slow but rather that regexes in general are slower than if you write a custom lower level parser. This is mostly due to the flexibility of regex and backtracking.

[omus]

Just needs a version bump, I assume we would want to bring along a few of the other CRs across the line before making a new release though.

For GitHub packages I've been bumping the version number outside of the PRs which avoids confusion on whether it has already been bumped. This will be 1.4.0 and I'll probably tag it soon after.

[mattBrzezinski]

Just needs a version bump, I assume we would want to bring along a few of the other CRs across the line before making a new release though.

For GitHub packages I've been bumping the version number outside of the PRs which avoids confusion on whether it has already been bumped. This will be 1.4.0 and I'll probably tag it soon after.

Yeah that's what I meant, @iamed2 converted me into this practice :P

[omus]

Quick aside. I benchmarked using possessive quantifiers to prevent backtracking and there was no performance improvement in this case

[iamed2]

IMO this should support ".." and we should continue to print "..". This is the standard notation for intervals in Julia.

[omus]

IMO this should support ".." and we should continue to print "..". This is the standard notation for intervals in Julia.

I'll make this change. I was starting to worry about how this could break tests which check printing.

[omus]

Parsing now supports both .. and , delimiters. I also improved quoting support by allowing an escaped quote character

[omus]

There's room for performance improvements here. I'll make a separate issue about that to be done later and get this feature out now

[omus]

Issue about improving performance: #131. The last round of regex changes didn't seem to negatively impact performance:

julia> @btime parse(Interval{Int}, "[1,2)")
  734.102 ns (13 allocations: 656 bytes)
Interval{Int64,Closed,Open}(1, 2)