use Perl

I want to stop C, C++ and Java people referring to their languages as 'type safe'. They're not 'type safe' - they're type constrained.

C has static types for two completely distinct reasons:

1. a partial attempt at Pascal / ADA style type bondage
2. as an efficiency - so the compiler need not generate code to handle dynamic types at runtime

In practice it's really hard to graft dynamic types onto a statically typed language. In fact once you've implemented dynamic types you've done so much work that you might as well implement the rest of the dynamic language. And if you're going to do that you might as well just use a dynamic language in the first place. The debatable virtue of type bondage is used to mask the fact that C-like languages can't get all funky and dynamic - even if they want to.

Type bondage sounds good on paper - "if we constrain this value really tightly at compile time we'll catch all sorts of runtime problems". That certainly sounds like it'd be a good thing.

Unfortunately it mitigates against a better thing: being able to write correct code that remains correct even when asked to process kinds of things that doesn't know about.

Dynamic typing? Don't need it!

You'll sometimes read something like dynamic typing offers no advantage in a properly designed program. The implication being that dynamic typing is for sloppy thinkers who can't plan ahead. That may be true. The problem is that at the limit we're all "sloppy thinkers who can't plan ahead".

The design of C does, in fact, acknowledge the need for dynamic typing. The sort() standard function achieves type-agnosticism by allowing the programmer to pass in

• the address of the things being sorted
• the number of things to being sorted
• the size of the things being sorted
• a function to compare the things being sorted

Whoever implemented sort understood that you'd want to able to sort just about anything. So, fortunately, they decided to make it type-unsafe, type-dangerous, type-scary. Its interface is necessary clunky - because C doesn't do dynamic types - but at least you can use it to sort your own data types.

We're invited to believe that sort is a special case, that in general you won't be needing that kind of flexibility. The necessity of type-inflexibility is spun as the virtue of type-safety. Safety is good! How can you oppose safety? Think of the children!

The current Perl QA wiki has been attracting a lot of spam recently - and we seem to have lost contact with Tyler MacDonald who as far as I know is the only person with admin access to it.

So I've moved all the content to perl-qa.hexten.net and set it up so that only registered users may edit. It's a shame to have to restrict access in that way; certainly we'll get fewer drive-by contributions as a result. On balance it's better than the message that a spam-riddled quality (quality!) wiki sends I think.

Please update bookmarks / links accordingly.

(warning: this item may contain traces of PHP)

A classic case of yak shaving: too much work on anyway, decided to take a break by turning our local community site into a wiki. Then I thought it'd be nice to embed a Google map showing a few local landmarks. Must be possible, right? Well it probably has been done already but it seemed pretty simple (and an interesting diversion) to write a MediaWiki extension to allow maps to be embedded.

So then you need to let users add points to the map - so the map needs button that spits out a chunk of markup that can be used to add a point. So I wrote some Javascript to do that.

Of course one extension is never enough (just say no kids). I was writing a load of code that'd be useful in other extensions anyway. What about a framework for extensions? Yup, the dreaded 'f' word. Oh well.

Anyway, to cut a long story short: Wiki Widgets.

One of the things I wanted to do (you know, once it became a project rather than a displacement activity) was to make widgets that, from the user's point of view, are discoverable. It's all very well installing an extension but if users can't find out about it they won't use it. Wiki Widgets has a pretty simple take on discoverability: make a blank page and add the markup:

<widget />

to it. When you save or preview the page you'll get a list of all the installed Wiki Widgets. For each Widget there's a link that takes you directly to a (wiki) page that explains how to use it. From there you should be able to select the widget you want. Wherever possible I plan to make the individual Widgets support additional discoverability. For example if you add an empty Google Gadget Widget like this:

<widget type="googlegadget" />

you should get a user interface that makes it easy to add the gadget you want and set up its parameters.

Anyway, back to the day job now.

Wow. I'm just installing Plagger on a couple of boxes. One of them is my regular laptop and it already has quite a lot of CPAN installed. Even so Plagger has what look like hundreds of unsatisfied dependencies.

I'm sure it's all going to be worth it - although all I want to do is merge a handful of RSS feeds. Right now it feels slightly like overkill for that task :)

I wish I'd never started this FizzBuzz thing now. Anyway, FizzBuzz/Perl in 60 bytes:

print$_%3*$_%5?$_:($_%3?Bu:$_%5?Fi:FizzBu).zz,"\n"for 1..100

Can it be bettered?

As you may know Perl's test result protocol is officially known as TAP - the Test Anything Protocol. Over on the perl-qa mailing list a discussion has been raging about the future of TAP. One of the points that came up was that TAP is no longer Perl specific. That means that a Perl mailing list is no longer the best place to discuss it. So that there's a language-neutral hangout for TAP fans I've created tap-l - a mailing list dedicated to TAP.

Most of the current TAP documentation lives on the Perl QA wiki. Again, not the best place if we want to persuade people that TAP is language neutral. So there's also a new TAP specific wiki:

http://testanything.org/wiki

FizzBuzz is an extremely basic competence test for programmers. So I thought I'd write a version extremely in BASIC. Here it is in BBC Basic's embedded 6502 assembler.

What's that got to do with Perl? Well I used my 6502 emulator written in Perl to develop it on.

   10 REM FizzBuzz in 6502 assembler
   20 DIM code% 1000
   30 OSWRCH = &FFEE
   40 OSNEWL = &FFE7
   50 work = &70
   60 DIM FizzM 4 : $FizzM = "zziF"
   70 DIM BuzzM 4 : $BuzzM = "zzuB"
   80 FOR pass% = 0 TO 3 STEP 3
   90 P%=code%
  100 [opt pass%
  110
  120 .FizzBuzz  LDA #1
  130            LDX #3
  140            LDY #5
  150 .FB1       SEC
  160            DEX
  170            BNE FB2
  180            JSR Fizz
  190            LDX #3
  200 .FB2       DEY
  210            BNE FB3
  220            JSR Buzz
  230            LDY #5
  240 .FB3       BCC FB4
  250            JSR PrDecimal
  260 .FB4       PHA
  270            JSR OSNEWL
  280            PLA
  290            CLC
  300            ADC #1
  310            CMP #101
  320            BCC FB1
  330            RTS
  340
  350 .Fizz      PHA
  360            LDX #3
  370 .Fizz1     LDA FizzM, X
  380            JSR OSWRCH
  390            DEX
  400            BPL Fizz1
  410            CLC
  420            PLA
  430            RTS
  440
  450 .Buzz      PHA
  460            LDY #3
  470 .Buzz1     LDA BuzzM, Y
  480            JSR OSWRCH
  490            DEY
  500            BPL Buzz1
  510            CLC
  520            PLA
  530            RTS
  540
  550 .PrDecimal STA work
  560            PHA
  570            TXA
  580            PHA
  590            LDA #0
  600            PHA
  610 .PrDec0    LDX #8
  620            LDA #0
  630 .PrDec1    ASL work
  640            ROL A
  650            CMP #10
  660            BCC PrDec2
  670            SBC #10
  680            INC work
  690 .PrDec2    DEX
  700            BNE PrDec1
  710            CLC
  720            ADC #ASC"0"
  730            PHA
  740            LDX work
  750            BNE PrDec0
  760 .PrDec3    PLA
  770            BEQ PrDec4
  780            JSR OSWRCH
  790            JMP PrDec3
  800 .PrDec4    PLA
  810            TAX
  820            PLA
  830            RTS
  840 ]
  850 NEXT

I've been working this week to add support YAML-based diagnostic syntax to TAP::Parser. Those changes haven't made it to CPAN yet but we'll probably make a release in the next few days. In the mean time you can find the bleeding edge version in the subversion repo.

Now TAP producers can elect to add machine readable diagnostics to test failures like this:

    TAP version 13
    plan 1..5
    ok 1
    ok 2
    not ok 3
    ---
    severity: fail
    message: "Array unsorted"
    data:
      got:
        - 1
        - 3
        - 2
      expected:
        - 1
        - 2
        - 3
    ...
    ok 4
    ok 5

The harness doesn't currently actually do anything with the diagnostic information but it does expose an interface that allows display plugins to be written. With the right plugin the test harness becomes responsible for displaying the detail of what went wrong. Currently that's something test scripts must do themselves, calling diag() to display a description of the problem. Because diag() outputs to STDERR the diagnostics can get out of sync with the test results.

Look out for more progress on display plugins and a practical end-to-end example of a test that emits structured diagnostics sometime next week.

From Adam Kennedy's YAML::Tiny:

    if ( $string =~ /^"((?:\\.|[^"])*)"$/ ) {
        my $str = $1;
        $str =~ s/\\"/"/g;
        $str =~ s/\\([never\\fartz]|x([0-9a-fA-F]{2}))/(length($1)>1)?pack("H2",$2):$UNESCAPES{$ 1}/gex;
        return $str;
    }

What, never? :)

This is a bit of a long shot. We have a test failure for TAP::Parser 0.51 on Windows:

http://www.nntp.perl.org/group/perl.cpan.testers/2007/03/msg437065.html

It's Perl 5.8.5 and to this (untutored) eye it looks as if it's not ActiveState's build. I've tried to build 5.8.5 here but can't get it to play nice with VC++ 8.0.

If anyone might happen to have 5.8.5 installed on a Windows box and could make test against both T::P 0.51 and a snapshot from the SVN and report back I'd be most grateful.

FWIW the tests all pass on Windows/5.8.8.