Use static regex for simple codespan results

[ashmaroli]

Rationale

It is wasteful to allocate /`/ and /`\Z/ every time a simple codespan is parsed. Instead use a static regex which is allocated only once each.

Context

Results from profiling Jekyll docs:

Before

Total allocated: 158.07 MB (1168216 objects)
Total retained:  10.3 MB (92547 objects)

After

Total allocated: 157.1 MB (1162835 objects)
Total retained:  10.3 MB (92547 objects)

[ashmaroli]

Profiler Diff Summary

--- master branch https://travis-ci.org/gettalong/kramdown/jobs/536192573
+++ PR branch     https://travis-ci.org/gettalong/kramdown/jobs/536323986

- Total allocated: 220.89 MB (1783975 objects)
+ Total allocated: 219.34 MB (1775370 objects)
  Total retained:  1.11 MB (1582 objects)

[ashmaroli]

@gettalong I guess we were wrong. The latest change isn't synonymous with the existing behavior:

irb(main):001:0> string = "`foobar`\n"
=> "`foobar`\n"

irb(main):002:0> string.sub(/`\Z/, '')
=> "`foobar\n"

irb(main):003:0> string.chomp('`')
=> "`foobar`\n"

irb(main):004:0> string.sub(/`\Z/, '') == string.chomp('`')
=> false

Shall I revert the latest commit or am I testing incorrectly?

[gettalong]

You are right, \Z matches before a new line at the end, \z matches at the end. I guess the original version it is.

[ashmaroli]

@gettalong Pinging you as a gentle reminder...

[gettalong]

@ashmaroli Sorry for the delay, I will find some time in the coming days for kramdown.

[gettalong]

@ashmaroli Even though the winnings are not much, the changes are not very invasive and I will therefore accept them.