Introducing Humane-re¶
OCaml is my favorite language, but one area where it (its tools rather) often falls short in practice is common string handling tasks where regular expressions are often involved. The kind of stuff that Awk and and scripting languages often get praised for. In other words, not getting in the way and allowing to get the job done with minimal boilerplate.
The story for trying to accomplish the same thing in OCaml is not nearly
as short. First, one usually looks at the String module and after a
quick scroll realizes that there’s no solution there. Second, the
Str
library is checked out. Realize that the interface is not user friendly
nor thread safe. If you’re a beginner, then you also start wandering
what kind of a functional language OCaml really is.
Luckily, now that we have OPAM, it’s much easier to look for solutions
beyond. If your string handling needs are simple enough then
core’s
or
batteries’
String module. If you’re lucky, your search ends there. Even if
you’re not, you still have plenty of good options:
pcre-ocaml Markus Mottl has written excellent bindings to the most popular flavor of regular expressions. In fact, I’d recommend these to anyone first if they’re writing an application or an internal library.
re2 Janestreet’s bindings to Google’s re2. The interface is quite nice but there’s a gajillion dependencies. Nevertheless, this is probably your best option if you’re looking for speed. As always, profile your code first.
ocaml-re This is the most interesting regex library because it’s written in OCaml. Its coolest feature is that it supports various regex syntaxes. Including: pcre, str, posix, glob, etc. In fact, it even has a drop in replacement for the builtin
Str. Unfortunately, re’s interface is rather prickly, especially for beginners. Fixing that problem is going to be the meat of this blog post. Nevertheless, this is what I’d recommend if you’re going to publish a library for others to use. It doesn’t force any non-ocaml dependencies on users.
Humane-re¶
Realizing that it’s too hard for users to do the right thing and use ocaml-re, I’ve created a little wrapper called Humane-re around ocaml-re that makes it easier to accomplish the common tasks. The goal is to cover 90% of the use cases with minimal incidental complexity.
For now, Humane-re is still an experiment so the interface isn’t stable
yet. I haven’t fleshed out the interface for replacement either, and
currently I’m only supporting the Str flavor of regular expressions.
Even with these limitations it’s already been useful for me. I’ll do a
few brief examples of how to do some common tasks.
To follow along, install humane-re with:
$ opam install humane-re
and load it in utop:
$ utop
# require "humane_re";;
# open Humane_re;;
Super Naive Email Validation¶
let is_valid_email =
let email_re = Str.regexp ".+@.+" in
let open Str.Infix in fun email ->
email =~ email_re
Extract all words¶
let extract_words = Str.(find_matches (regexp "\\b\\([A-Za-z]+\\)\\b"))
Parsing HTTP Header Like Value¶
let parse_header =
let re = Str.regexp ":[ \t]*" in
fun header ->
match Str.split ~max:2 re header with
| [name; value] -> Some (name, value)
| _ -> None
I’ll admit, for these simple (but not contrived!) examples there’s no
great improvement in readibility over Str. At least we’re not
relying on any global variables. However, humane-re pulls ahead of
Str in readability when groups are involved. I’ll show how to use
groups in the next example.
Extracting Links Matching a Predicate¶
Where the predicate here is links to a certain website (e.g. imgur). Don’t ever use regular expressions to parse HTML in practice. This is only for demonstration purposes:
let extract_imgur_links page =
let is_imgur s =
let open Str.Infix in
s =~ (Str.regexp ".+\\bimgur\\.com.+")
in
let re = Str.regexp "<a href=\"\\([^\"]+\\)\">\\([^<>]+\\)</a>" in
page
|> Str.fold_left_groups re ~init:[]
~f:(fun acc g ->
match Str.Group.all g with
| [href; text] when is_imgur href -> (href, text)::acc
| _ -> acc)
|> List.rev
The whole interface is contained in S.mli. I could reproduce it here but it will just go out of date. The ocamldoc isn’t there yet but the interface should be straight forward enough. Once again, send me suggestions, questions, critique, etc.
What’s next?¶
At this point I’m trying to collect as much feedback as possible about the interface because providing a nice interface is the first goal of this library. In particular, an interface for substitution would be very welcome.
The second goal is to support the different ways ocaml-re allows you to
construct regular expressions. I’m not very fond of Str’s regex syntax,
but it does have the practical purpose of allowing me to port old
Str code. The third and lofty goal is to implement humane-re’s
interface with other backends. There’s probably some value in not having
to commit to code to any particular regex implementation (aside from
benchmarking purposes ;D).