commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Henri Yandell <flame...@gmail.com>
Subject Re: Proposed Contribution to Apache Commons,
Date Fri, 02 Oct 2015 06:58:30 GMT
On Tue, Sep 29, 2015 at 5:42 PM, Phil Steitz <phil.steitz@gmail.com> wrote:

> On 9/29/15 3:55 PM, Gary Gregory wrote:
> > Norman,
> >
> > Hello and welcome to Apache Commons.
> >
> > It's not clear to me why Naomi is better than regular expressions.
> Pointing
> > to Javadocs is not the best way to get traction.
> >
> > Your project would be better served by having some documentation on your
> > front page with an example driven tutorial.
> >
> > Is Naomi faster than REs?
> >
> > What can I do in Naomi that REs can't do? And vice-versa.
> >
> > Examples of this on your front page would help you at least get folks to
> > consider learning a brand new way of doing things...
>
> +1
> The code in SimpleExamples starts to get to this.  Looks interesting
> and powerful.  Either here or on the github readme you should take a
> stab at explaining a little more how hard problems using regex get
> easier with naomi, illustrated with some simple examples.  Then
> maybe with help from community members here, you can develop some
> overview / getting started docs that help people get into the code.
>

+1.

Reading SimpleExamples, my summary would be a boilerplate description of
"It replaces the arcane regular expression language with an API". It
reminds me of command line argument parsers. Perl had/has a great regular
expression like command line argument parser, but it was cryptic and you
either loved it or hated it. Then along came Commons CLI, args4j and all
the others, providing a more OO/procedural API instead of its own mini
language. Not as 'powerful' (in that you had to type more), but simpler (in
that you didn't have to learn a new lingo and didn't have to juggle
multiple languages inside one context (a source file)).

I definitely need that user manual. It's hard, with a brain trained on
regular expressions, to read 'Pattern greek3=new CharSequencePattern("?")'
and realize (I think) that it means a literal ? character. It's also the
primary way it'll be successful. You need that educational path that
explains what a ExplicitCharClass is for, rather than randomly clicking on
javadoc :)

There'll also be much debate to be had I suspect. Is "a-e" too complex,
compared to "abcde" or "a","e". Which parts of regex are worth supporting,
vs not. Can I mix bits of regexp with bits of Naomi?   new
ExplicitCharClass("a-eg-p").

Random.... I'd like the idea of varargs for automatic and'ing. ie:

new ExplicitCharClass("a-p", "!f")   [and is a not char class too complex?].

Continuing on my summary, as I peruse the code a little more, I'd go with:

"Build a regular expression via an API, not an arcane language of its own".

I'd love to see that grow to:

"Express regular expressions as objects, or mix and match objects with that
arcane mini language we all love or loathe".

Hen

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message