Add multi-line support

[tustvold]

Fixes #40

[codecov]

Codecov Report

Merging #63 into master will increase coverage by 1.04%.
The diff coverage is 93.75%.

@@            Coverage Diff             @@
##           master      #63      +/-   ##
==========================================
+ Coverage   86.77%   87.82%   +1.04%     
==========================================
  Files           7        7              
  Lines         242      271      +29     
==========================================
+ Hits          210      238      +28     
- Misses         32       33       +1     

Impacted Files	Coverage Δ
dotenv/src/iter.rs	93.87% <93.75%> (-0.86%)	⬇️
dotenv/src/find.rs	90.90% <0.00%> (+3.95%)	⬆️

---

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 3c1a77b...0b99e5b. Read the comment docs.

[tustvold]

The parser on master doesn't handle lines with unterminated ' or ", except where enclosed in a terminated block of the other character

For example:

KEY=asd'f => ERROR
KEY=asd'f' => asdf
KEY=asd'f'" => ERROR
KEY=asd'f'"a" => asdfa
KEY=asd'f"' => asdf"
KEY=1'2'3' => ERROR
KEY=1'2'3'4' => 1234

So this shouldn't break anything that was previously working

The implementation provided in this PR won't handle cases such as below, but I'm not sure people would expect it to

KEY='sdfsdf'"sdf
sdfsd"

On a related note, whilst experimenting I found that the parser on master doesn't handle escapes within ' which might be a bug?

KEY='\'' => ERROR
KEY="\"" => "

[pksunkara]

Looks great. Very reasonable solution too!

[daboross]

This looks to me like it would cause an exponential slowdown if a quote contained a large number of lines - it would have to loop through every character read so far for each new line.

Would it be possible to make this a stateful parser which keeps track of what quotes have been seen so far, and only looks at each character once be possible without making it significantly more complicated?

I mean, the current code won't slow down any current use cases, and env variables are unlikely to be very large. But it seems non-ideal to have an exponential slowdown if it's avoidable?

I'm asking as an observer: I don't have any ownership over dotenv.

[hoijui]

I'll have a go at it

[hoijui]

have it working:
#78

[hoijui]

superseded by allan2#3