Earlier today I outlined the goals of my current release. One of the problems I was looking to fix was with the function, getMaxEntries(). The job of this function was to create an xmlHttpRequest() in order to retrieve the maximum number of push entries in the repository. The pushlog displays data in reverse chronological order and thus I need to know the maximum number of entries so that I know which entries to display first to maintain the same order.

For example in my test repository I have 2613 total entries. The first 10 entries are displayed by default so when the user scrolls down to load more data the very first entry shown would be #2603, then 2602 and so on. Previously, to retrieve the maximum number of entries I was using the following function:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

function getMaxEntries() {
  var entries = new XMLHttpRequest();
  var max = 0;
  entries.open('GET', '/json-pushes?startID=0&endID=1', true);
  entries.onreadystatechange = function() {
  if(entries.readyState == 4)  {
    if(entries.status != 404) {
      var entryData = JSON.parse(entries.responseText);
      max = entryData[1].max;
 
      start = max - 10;
    }
  } else 
      return 0;
  } 
  entries.send(null);
}

I didn’t really like this method of using an xmlHttpRequest() just to retrieve one value. I had to call this function onPageLoad() to calculate the maximum number of entries. This had to be done before my OnScroll function was called or various horrible errors would occur.

I’ve noticed something very unique about JavaScript. It doesn’t wait for one function to finish executing before going on to the next function. If your first function is taking too long, JavaScript will move on to your next function. Now, this feature was causing me quite a few headaches. As I said before, getMaxEntries() had to finish executing completely before the OnScroll function was called, otherwise things would go horribly wrong. getMaxEntries() sets the value of start, which is then used by loadEntries() to retrieve data (called OnScroll). Since getMaxEntries() was taking too long to execute, JavaScript was moving on to loadEntries() without setting the value of start, completely wrecking my logic.

Also, getMaxEntries() was causing Firefox to freeze for 10-15secs for reasons unknown to me. This way of calculating the max number of entries wasn’t very elegant and totally unacceptable. I needed to come up with a better solution.

The Solution

hgweb uses a template system called genshi. Basically, the maximum number of entries is calculated via a database query on the server side. In order to pass this server side variable to the client side I had to add the line 4 to pushloghtml() in hgpoller\pushlog-feed.py:

1
2
3
4
5
6
7
8
9
10

    return tmpl('pushlog',
                changenav=changenav(),
                rev=0,
                max=query.totalentries,
                entries=lambda **x: changelist(limit=0,**x),
                latestentry=lambda **x: changelist(limit=1,**x),
                startdate='startdate' in req.form and req.form['startdate'][0] or '1 week ago',
                enddate='enddate' in req.form and req.form['enddate'][0] or 'now',
                querydescription=query.description(),
                archives=web.archivelist("tip"))

Now, this change ^ allowed me access to this variable on the client side. All I had to do was to use this format: {<var_name>} or #<var_name>#. However, one draw back as far as I know is that I can only access this varialble in HTML, not in JavaScript, which is a significant drawback.

So in order to retrieve this data I added an id attribute the following div tag in hg_templates\gitweb_mozilla\pushlog.tmpl:

1

<div id="#max#" class="page_header">

Now, I could set the value of start to the maximum number of entries:

1

start = $("div").attr("id");

Thus, I can easily pass the correct value of start to loadEntries(), which makes an xmlHttpRequest() to retrieve the correctly ordered data when the user scrolls down.