WeirdLooking.com: Michael Barton's Blog

barcode scanning

2006-12-06T09:11:07Z

You may remember I was playing around with writing a Flash/webcam barcode scanner a while back. Since nothing has ever come of that project, I decided I’d go ahead and stick it somewhere for a while with a quick and dirty application I wrote in about a half hour (and it shows!).

What happens is you scan a barcode, and it hits up Amazon for some information about the item via their ECS webservice and displays it. Their service returns a different XML format for each category, and I haven’t bothered to account for them all. You’ll probably have the best luck scanning books and CDs.

broken gmail feeds

2006-09-04T18:53:29Z

Gmail provides Atom feeds of your inbox, but they come at a price. And the price is the possibility of having to deal with them. Mine kept causing an error in RSS Bandit. At first I tried to blame the aggregator, but then I took a closer look at the feed itself…

<entry>
  ...
  <modified>2006-09-03T24:56:06Z</modified>
  <issued>2006-09-03T24:56:06Z</issued>
  ...
</entry>

That email was received at nearly 25 o’clock, eh? The interesting part is, it came in at 1am on the third (GMT). So not only is it an invalid date, it’s off by 24 hours. I guess they just forgot to mod the hour by 24 or something.

I submitted a bug to Google, and they promptly replied with an automated message! It seems they aren’t able to respond directly to my report, but they are working on these issues. I know I feel better about it ever getting fixed.

So what I did was, I built myself a little proxy script that fixes the time. I still need to make it pass through the cache headers, but I’m a bit lazy for that right now. See, they’re forcing me to waste their bandwidth.

<?php
  if (!isset($_SERVER['PHP_AUTH_USER']))
  {
    header('WWW-Authenticate: Basic realm="My Realm"');
    header('HTTP/1.0 401 Unauthorized');
    echo 'You are NOT authorized.';
  }
  else
  {
    header('Content-type: text/xml');
    $ch = curl_init("https://mail.google.com/mail/feed/atom");
    curl_setopt($ch, CURLOPT_UNRESTRICTED_AUTH, true);
    curl_setopt($ch, CURLOPT_HTTPAUTH, CURLAUTH_ANY);
    curl_setopt($ch, CURLOPT_USERPWD, $_SERVER['PHP_AUTH_USER'].':'.$_SERVER['PHP_AUTH_PW']);
    curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
    echo preg_replace('/(\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}Z)/e',
      'str_replace(\'T24:\', \'T01:\', \'$1\')', curl_exec($ch));
  }
?>

i hate actionscript

2006-07-03T11:24:35Z

I needed to make a Flash movie containing pictures that could be dragged out onto the surrounding HTML elements. And it almost worked, too.

You can pass information from Flash to HTML by sending the browser to a javascript: URL, so the idea was to trap the mouse-down event in Flash, then pass a message to the javascript telling it that it’s currently in a dragging state. Replicate the dragging graphic in both Flash and HTML, and you can almost imagine this method working.

Except… if you try to drag something outside of Flash, it refuses to give the mouse events back. Flash continues receiving events though, so you could theoretically pass those back to the browser yourself. But calling getURL(’javascript:…’) to share the current mouse position causes IE to “click!” for every pixel the mouse moves.

I’m not really a Flash guy, so this may be possible via some other chicanery. But based on currently available data, I’m going to call it a defeat.

Mike: 0, Flash: 1

javascript mud client 2

2006-03-28T17:04:14Z

I woke up early this morning and decided to start from scratch on my javascript mud client. After writing the single worst implementation of a web server ever (it would have gone nicely with my networking final project, which included the single worst implementation of a web client ever), I was able to get it working.

Basically, I rewrote the MUD to accept HTTP requests and wrap a javascript call around the output from each pulse. Taking Matt's advice, I decided to use <b> tags with short class names, rather than the <span style=”color:…”> stuff I was doing before.

Example (mud prompt): <script>parent.push('<br><b class=g><20hp 100m 100mv> <b class=X>');</script>

The browser downloads this “page” inside a hidden <iframe> (using the hack I previously posted for Internet Explorer, just a plain iframe for FF). The push() function in the frame’s parent creates a new <pre> tag, stuffs the argument text inside it, then appends it to the display <div>. Then the div’s scroll is set to the bottom.

Mud input uses a form submit handler, which takes the value of the <input> and sticks it into an XMLHttpRequest, which is sent to the aforementioned worst web server implementation ever.

I hacked it into QuickMUD rather than my own code base so that, should I ever decide to clean it up, other people can use it.

Okay, so it’s pretty trivial. But it was fun.

web-based mud client?

2006-03-26T06:11:26Z

This blog post about Google’s web-based instant messaging is making the rounds:
http://alex.dojotoolkit.org/?p=538

So here’s my thinking… If I can open up a second port on my mud and run a simple HTTP server on it (lots of people have done this already), I can pretty easily hack write_to_descriptor to send information to a persistent HTTP connection. In fact, it’ll practically be a regular mud connection except color codes and newlines and stuff will be parsed into their equivalent HTML (fairly straight-forward modification of the color code). And interpret will be called from an HTTP post. Then some work in javascript and I can have a web-based mud client.

I’ll have the first Web 2.0 mud.

i don't like (comment) spam!

2005-12-23T06:06:20Z

You know, all of the comment spam is coming from the servers of Web hosting companies. Which makes sense; they rarely monitor or charge for outgoing connections. I assume that this page made its way into an automated comment spamming tool or something. I’m going to delete most of that spam, I think. Then I can remove the lame rel=”nofollow”s on comment links.

As mentioned, I assume that this spam is posted by some sort of automated tool. So, I have put an end to automated posting. How, you ask? I just hid one of those evil Unix timestamps in the form. If you post a comment 30 minutes after that timestamp, it gives you an error (and a preview where you can click Submit again, so it’s not evil).

Yeah, that’s easy to circumvent. The form also includes a checksum of the time (generated via fantastic, super-secret cryptographic methods that you’ll never figure out, so don’t even bother trying). The end result is that stale instances of the form, like those I assume live in a spammer’s database, won’t work.

They could still easily download the page and scrape out a valid timestamp and checksum, but I think it’s probably enough for now to not be the low-hanging fruit.

human-readable date differences

2005-12-21T11:29:32Z

I had an idea today. I wanted to include human-readable ages beside some of the dates on my blog (e.g. “1 day, 8 hours ago”). I searched around a bit, and all of the solutions I found just divide the difference in seconds by 31536000 to get years and so on. It doesn’t take leap years or different-sized months into account.

Today is December 21, 2005. To me, 2 years, 2 months ago was October 21, 2003.

It took a while to write. The first version used brute force to get the difference out of mktime. That worked perfectly but was slow. I used it to generate a bunch of test cases, though.

What I eventually figured out is that you can just subtract the beginning years, months, hours, etc. from the current ones. After that, use borrowing similar to long subtraction to get rid of any negative numbers. Seconds borrow 60 from a minute, hours borrow 24 from a day, days borrow from an array of month lengths. You get the idea. Anyway, here’s the code, which I’ll post on the SmartyPlugins wiki as soon as I figure out how.

<?php

/**
 * Smarty plugin
 * -----------------------------------------
 * File:    modifier.human_age.php
 * Type:    modifier
 * Name:    human_age
 * Params:  from_time    unix timestamp (required)
 *          to_time      unix timestamp (default=false (means now))
 *          strip_zeroes remove units with zero values ("0 days") (default=true)
 *          limit        limit the number of units returned (default=no limit)
 * Author:  Michael Barton (http://www.weirdlooking.com/email)
 * Purpose: Returns a human-readable difference between two times
 *          e.g. "1 day, 8 hours, 48 minutes, 57 seconds"
 * Remarks: Correct for leap years and different-sized months.
 *          Well, my idea of correct.  Turns out it's slightly subjective.
 *
 * @link http://www.weirdlooking.com/blog/45
 * -----------------------------------------
 */
  require_once $smarty->_get_plugin_filepath('shared','make_timestamp');

  function smarty_modifier_human_age($from_time, $to_time = false, $strip_zeroes = true, $limit = false)
  {
    if ($to_time === false)
      $to_time = time();
    $order = array('year', 'mon', 'mday', 'hours', 'minutes', 'seconds');
    $names = array('year', 'month', 'week', 'day', 'hour', 'minute', 'second');
    $sizes = array(12, 2=>24, 60, 60);
    $month_lengths = array(31, 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30);
    $from = getdate($from_time);
    $to = getdate($to_time);
    $delta = array();
    foreach ($order as $unit)
      $delta[] = $to[$unit] - $from[$unit];
    for ($i = 5; $i >= 0; $i--)
      if ($delta[$i] < 0)
      {
        $delta[$i-1]--;
        if ($i == 2 && ($to['mon'] == 3 && !($to['year'] % 4)))
          $delta[$i] += 29; // fix this before the year 2100!!!
        else if ($i == 2)
          $delta[$i] +=  $month_lengths[$to['mon']-1];
        else
          $delta[$i] += $sizes[$i-1];
      }
    $delta[2] -= ($week = floor($delta[2]/7)) * 7;
    $delta = array($delta[0], $delta[1], $week, $delta[2], $delta[3], $delta[4], $delta[5]);
    $ret = array();
    foreach ($delta as $i => $val)
      if (($delta[$i] || !$strip_zeroes) && ($limit === false || sizeof($ret) < $limit))
        $ret[] = $delta[$i].' '.$names[$i].($delta[$i] == 1 ? '' : 's');
    return join(', ', $ret);
  }
?>

mike:1 coldfusion: 0

2005-12-20T01:58:49Z

Cole mentioned that he was trying to parse Unix time stamps from Yahoo’s traffic web services to something human-readable. In ColdFusion. In case you don’t know, CF runs on top of Java and has access to its entire API. “Easy”, thought I, Java’s Date object can deal with milliseconds since epoch, very similar to your Unix time stamps in most regards. ColdFusion’s DateFormat function can then be used to translate Date objects into a human-readable string.

<cfset time = 1135038970>
<cfset date = createObject("java", "java.util.Date").init(time * 1000)>

Nope. The constructor takes a long argument and coldfusion’s scalars cast to doubles when sent to java. So we turn to ColdFusion’s answer, JavaCast.

<cfset date = createObject("java", "java.util.Date").init(JavaCast("long", time * 1000))>

This is where it got weird. Coldfusion threw an error that the number was too big to be cast to an int. What?! Why is it casting anything to an int?!
I tried a slight variation:

<cfset date = createObject("java", "java.util.Date").init(JavaCast("long", time) * 1000)>

It cast time to a long alright, but then multiplying it by 1000 casts it back to a double. So I wracked my brain for a way to get this number multiplied by 1000 into a long. Laughing maniacally, I typed it out:

<cfset date = createObject("java", "java.util.Date").init(createObject("java", "java.lang.Long").parseLong(time & "000"))>

You proved a worthy adversary, ColdFusion, but your crane style was no match for my tiger style.
Oh, and there you go, Cole. :D