Hi all,
I've pushed a new branch no-undef-hash-values with a new experimental
option for the Perl encoder. This will cause the encoder to forget
every hash key that is associated to an undefined value. This
dangerous option is to allow micro-optimisations for the small number
of cases where we don't care distinguishing between existing keys with
an undef value and non-existing keys. (Real-world testing with a
random production data blob over here shows a 5% saving in size, but
obviously ymmv.) I'm not sure it can be useful / wanted yet, so
comments welcome.
There's one commit, and it's rather simple:
https://github.com/Sereal/Sereal/commit/6966503830ca7f919847f3b1866b5ac02f051250