2015-06-14 Go Oddmuse

Sometimes I wonder whether I should move Oddmuse away from Perl 5 to something else. Something I would like to learn. An opportunity to redo everything, from scratch. Not many people are using Oddmuse, so I might as well write something else, for myself.

Oddmuse

I recently read Writing Web Applications for Go, where they develop a simple wiki. I wondered about storing the data in a git repository. I found git2go. Hm...

Writing Web Applications

git2go

​#Oddmuse ​#Web ​#Go

Comments

(Please contact me if you want to remove your comment.)

⁂

Perl 6 is an obvious candidate. And yes, they did not just change **print ’x’** to **print(’x’)** like they did in Python, it is a completely new language that you have to learn.

Speaking of Go, there is a great video called Perl 6 for Mere Mortals, and here is a direct link to a part of it which is somehow 😄 related to Go. That’s not a meaningful comparison, but it should give you the idea. You can also watch that video from the beginning, it is very interesting if you have not investigated into Perl 6 yet.

Perl 6 for Mere Mortals

Now, if we think about all of the languages out there, what are the actual features that are required for Oddmuse?

Well, since Oddmuse is all about parsing the wiki text, you need some tools for parsing. Perl 5 has regexes (um, okay), Perl 6 has grammars (exactly what we need!), and Go has something as well (like this, which is, again, verbose as hell), but I’m not an expert. You can also use libs, but having a built-in support is probably advantageous.

this

Somehow I can’t really think of any other criteria, everything else probably exists in every other language out there. You might also say “performance” but the only website suffering from that is probably emacswiki, and it all depends on your actual goals – for example, better execution time can be achieved by using concurrency, which does not really lead to better CPU time. Concurrency in Perl 6 – yes (but maybe just a bit flaky at the moment), performance in Perl 6 – well, not now, but the potential is there (i.e. gradual typing).

In other words: what is the motivation? “Just to learn” is not going to get you anywhere, not even make you learn something (learning to find reasons for doing stuff (i.e. causation) is part of the learning, doing stuff purposelessly is harmful to the thinking process).

– AlexDaniel 2015-06-14 14:05 UTC

AlexDaniel

---

Perhaps Go just makes me want to delve into the low level stuff. I like the post about Handwritten Parsers & Lexers in Go. 😄

Handwritten Parsers & Lexers in Go

I keep thinking about all the design decisions. Filenames are pagenames. Namespaces are subdirectories. Can we switch to “git first”? No more log files. We no longer need keep files. How would modules work? Would you simply recompile the wiki? Dynamic linking? Would I want to use a templating architecture? People seem to like that. Was “printing as we go” a good decision? Perhaps it no longer matters much.

I need to think about extensibility of Cajun, the Wiki Creole lexer and parser.

Cajun

– Alex Schroeder

---

These are very good questions. Maybe you should write down these ideas on Oddmuse: Revolutionary Changes.

Oddmuse: Revolutionary Changes

• “git first” – Yea, but... This would effectively break the goal of having one script file that just works (and I think that this is more important than having no recent changes and keep files logic in the core).
• “How would modules work? Would you simply recompile the wiki?” – this would effectively throw away all of the nice things like Module Updater or Module Bisect (or at least make them much harder to implement). It will also break the potential of a few other ideas that keep floating in my head. In other words, “simply recompile the wiki” is not that simple.
• “Would I want to use a templating architecture? People seem to like that.” Yea, but... It seems like if we switch to Perl6 then the core will collapse into a small piece of code with all of the irrelevant stuff separated into grammars or other classes, or maybe just tiny subroutines. Which brings us to the next point...
• “Was “printing as we go” a good decision? Perhaps it no longer matters much.” – Good decision, but it really depends on your understanding of “as we go”. For example, what if we parse the wiki text first, before printing anything except the header? Then, once we start processing the parsed text, we can print stuff as we go. That’s exactly how it will work in perl6 (at least, that’s the easiest way) – first you slam the wiki text into your grammar and wait for it to be processed (oh by the way, while this is happening you can actually do other stuff asynchronously, like start reading the tag index or whatever). Then, when your wikitext is parsed, you can start deciding what to print now. And this is *so much better than what we have now*. Want to print table of contents? Just take the required information from the match object, no need for dirty hacks. In other words, in perl6 it is very easy to have parsing and actions separately, which is beneficial in many ways. But at the same time, once you start processing the actions you can start printing as you go (which makes sense, what if there is some complex search stuff included on the page, you’d probably want to see at least half of the page while this is going).
• - AlexDaniel 2015-06-16 00:22 UTC
• —

Module Updater

Module Bisect

AlexDaniel

And yeah, if you are seriously thinking about Go, what if we make a comparison table with Perl6 and Go, and the relevant language features? Both of the languages have some interesting stuff to offer, for example Perl6 to JS and Go to JS. And there are definitely some winning points for Go, especially associated with execution time and maturity.

Perl6 to JS

Go to JS

– AlexDaniel 2015-06-16 00:31 UTC

AlexDaniel

---

Hehe, I think you and Oddmuse: Revolutionary Changes is what started this all! I looked at Oddmuse as a “mission accomplished” project. With the migration of Emacs Wiki I’ve realized that a big site might want to serve its own pages without running a web server. I tried to understand PSGI/Plack in order to have Perl run Oddmuse like a web application. When I saw the wiki tutorial for Go, and how I got a web server up and running in no time, and how easy it was to include some basic markup, I was thrilled! Sure, the code doesn’t have a lot of features. But it beats the SmallestWiki candidates. 😄

Oddmuse: Revolutionary Changes

PSGI/Plack

the code

SmallestWiki

You arguments in favor of some of the core architecture decisions made for Oddmuse are good – and your hopes for Perl 6 and improved parsing are infectious! I’m very much looking forward to Perl 6.

– Alex Schroeder 2015-06-16 06:02 UTC

---

It’s also cool to look at the TinyWiki source code again and rediscover why I rewrote the Usemod Wiki style search and replace code with the extensible state machine code we have right now.

the TinyWiki source code

– Alex Schroeder 2015-06-16 08:00 UTC