development corrupts

Looks like I'm getting a patch into Mercurial :)

Ta, patch is in stable

I paid the (one-time thankfully) $5 fee, and have submitted the Copy Link Text Chrome extension to the gallery. Happy to have received some stars \o/

I've also created a git-hub repository for the extension.


I've been trying to spend some of my free time doing open source stuff that's not DW, so I can see what's out there (maybe introduce some things back into here, maybe not!).

So far, it's been fun. My biggest, er, thing, is that I apologize a lot. I'm trying to turn that around into thanks, so instead of saying, "I'm sorry I got my extension idea from a similar Firefox extension", I'm trying to say, "I loved that extension, I think it's fantastic, thank you for it! (In homage thereof I have written one for Chrome too *g*)"

Or, instead of saying, "Sorry for grabbing your lovely icon from the icon set you're offering for free", I'm forcing myself to see that "Gorgeous icon set, great icon that fits my needs exactly, thank you :DD" is the more appropriate response.

And instead of saying, "Hey, I noticed this issue, and I wrote a patch; the github documentation says to use forks for this, I don't know if it's right or if I seem rude. I'm sorry for even bringing it up x_x", I have a message which basically goes, "Hi, I loved $extension. There's just one little thing that makes it less than perfect, and it's $issue_I_ran_into. *contributes patch upstream*"

THAT KINDA THING.

Anyway, it's ridiculously hard to not apologize all the time. Ridiculously ridiculously hard! But I'm trying to turn it around, because while it feels inappropriate not to apologize (it really really does), it feels even more inappropriate to apologize. Before, I'd have just decided it was not worth the risk of offending someone by being awkward, and just backed away without ever having said anything. Now, I'm forcing myself to acknowledge that hey these people shared these things for a reason, and the polite thing to do is to use them ;-) (And give back too)

AND THAT IS THE REALLY LONG WINDED STORY BEHIND WHY I AM PUTTING MY EXTENSION ON GITHUB.

Have finally redownloaded Google Chrome because I need it for some troubleshooting. Before doing so, I looked up how to disable auto-update, and discovered that it's now a matter of changing a setting via the Terminal, rather than deleting the auto-update program. This seemed a lot less virus-like creepy, so I have Google Chrome on my system right now, and feel okay about it.

While moving over settings, I discovered I'd have to restart Firefox -- shutting down Firefox took over an hour! And ate up all my CPU! I had to force-quit it after all; luckily didn't lose any data.

As a bit of amusement, I decided to see what it would take to get me to switch over to use Google Chrome regularly. At a minimum, I'd need:

• Adblock and/or JS blocker (I went with Adblock, but I'll try to retain all my casual browsing on Firefox with NoScript on)
  
• Something to keep count of open tabs (settled on TabGlutton)
  
• Ability to quickly edit and preview CSS (probably somewhere, but I haven't looked for it; just the current hacky ways are enough for now)
  
• Firebug or equivalent (built-in)
  
• Greasemonkey script support (built-in)
  
• ReadItLater (Found ChromeItLater; slightly clumsier than the official extension Firefox, but workable)
  
• A JS shell (I have a bookmarklet, but I eventually decided to go with the jsshell extension)
  
• a JSON prettifier (JSONView for chrome)
  
• The ability to copy just the link text instead of the link address

I found what I needed, or a good-enough equivalent, for everything but the last one. Searching through the extensions site and through the greater web turned up nothing. So.... I ended up writing my own.

( screenshot )

It didn't take long; however, figuring out the difference between the content_scripts side (which lets you manipulate the webpage) and the main extension body in background_html (which lets you manipulate the browser behavior) took some doing.

The basic "hello world" example shows you how to create a regular extension, one which interacts with the browser. It took me an embarrassingly long time to figure out there was something beyond that.

After I figured out that the reason my extension wasn't doing anything was because it was running once when the extension loaded, when it needed to be running once per page, I slowly came to the realization that I'd need something like body.onload. Unfortunately, even after going through all the links from the getting started portion of the tutorial, I had still overlooked the other part of the API which would enable me to interact on a per-page basis (hello content_scripts!). I tried to force it by using chrome.tabs.onUpdated.AddListener... if ( changeInfo.status == "complete" ) as a poor extension writer's body.onload and that worked, but wow was it ugly :D

I can't remember how I found the link, but after finding that section in the documentation, I was more than halfway there.

In my initial version, I tried to programmatically create a range (document.createRange()) to select the text in the content scripts portion of the page. Right after that, I'd rely on some timing hackery to run document.execCommand("copy") (yeah I didn't know this function existed until now either) in the extension trusted area. It worked, ish, but was ugly and hacky and brittle.

Reason I had to put document.execCommand in the extensions area: it seems to need to be run with higher privileges. If it were available to the website, any website could see your clipboard, and that could potentially leak information. Since extensions are more trustworthy -- you choose whether to install them, and I believe they're vetted to some extent by reviewers before they are distributed in the gallery, they are allowed to execute with higher privileges than Joe Random Website.


On the second version, I figured out I could send just the relevant information (the link text) to the main extension body via a communication pipe, and have the copying take place there. That took away all the worst of the brittleness and most of the hackishness (there still is some ;-)), as well as gave me some flexibility for being able to output in multiple formats should I choose to code that in in the future, and I am quite happy with the results.

I can't take credit for it though: I had the glimmerings of the idea on my own, but I didn't actually get it to work until I peeked into the source code of Auto Copy.

So anyway, there it goes. Have made a Google Chrome extension, it wasn't very painful once I managed to wrap my mind around how everything fit together, and I'm happy with the results.


Ohh one final thing: for a long time, I thought that console.log() didn't work in extensions, and struggled with that. It turns out:

a.) console.dir in Google Chrome is the equivalent of console.log in Firebug; and
b.) If your console.log calls are in your main extension body, you need to open the background.html view from the extensions tab, as the console log output is put there. If your console.log calls are in a content_scripts js file, then they'll show up in the webpage

And there! :D

Have been fighting Opera, trying to make it understand that I need the tab and enter keys for autocomplete. Lots of googling later, finally realized that I could ignore all the talk of trying to bind to the keypress event or whatever, and instead use the existing hooks to set a "just completed" flag, after an item has been selected. Then, if the form gets submitted by the enter key, or if we leave the text box, check whether we'd just completed something.

It sounds horribly complicated, but it's only three tiny chunks of code (or it was, after I got rid of all the other more complicated, but unnecessary, approaches \o/)

Really useful article about jquery ui widgets:
* Understanding jQuery UI widgets: A tutorial was crucial to my understanding of events in jQuery.

Also, jsFiddle is awesome. It's a combination of pastebin and an active dev environment. I talk about it a bit in the javascript community, but really it's much better than I can make it sound *g*

(A couple of weeks ago, I posted a link to Google Playground. This is like that, but even niftier if you just want to play with JS and JS frameworks, not Google's APIs).

In my attempt to conquer Bug 2344: Default View/Default filters should be default when adding from hover menu, I wrote this line of code:

$success ||= $filter->add_row( userid => $u->id )

What I thought it said:

* add this user to the filter
* once you've added the user, check whether it was successful, and keep a running tab on the status so you'll know in the end whether everything succeeded or not

What it was actually saying:
* once you've successfully done one thing (e.g., added the user to a filter or subscribed to that user), we're done. We don't need to do anything else (in this case, adding a user to the filter).

So now I've rewritten it more properly as:

$success = $filter->add_row( userid => $targetu->userid ) && $success;

(I got the boolean logic initially wrong too *rueful*)


This was an entire afternoon's worth of frustration (whyyyyyyy isn't it adding the user to the filter? Why doesn't it print out any of my warn statements within add_row? I didn't realize I wasn't calling it at all), solved with one of those flashes of insight you get when you get up and grab a glass of water and happen to reread your code when you get back.

I'm doing Bug 216: Renaming, and maaaaaan.

It's kind of funny -- I studiously ignored most of the renaming support requests back on LJ, because it seemed complicated and I didn't want to dig into the code, and now here I am, writing similar logic from scratch *snickers*

I'm having great fun. This seemed like one of the projects where tests would have a great effect, so I spent some time figuring out how to write tests from scratch, rather than just trying to tweak existing tests.

My tests look kinda like this:
( tests )

I'm coding by first writing a bunch of tests, then writing code to make the tests pass. Every time I think of something new I should check for, I dash off a quick TODO block and then go back to writing either my test or my module. That helps me not lose my train of thought, but also means that I don't lose my ideas.

I'm starting with a small set of limited functionality, and then gradually expanding the scope. I actually started by disallowing everything except user-to-unregistered, and now I'm relaxing the rules as I add more. So far, I have personal-user-to-unregistered, with redirect and without redirect, and lots of checking for whether $fromu can rename to $tousername.

I'm in the middle of user-to-other-user, where the second user is just moved out of the way.
I estimate that I'm only about 10% of the way through -- I still need community-to-unregistered, community-to-other-community (or journal??), two-username-swap, keeping or removing various links -- subscriptions and the like, then catching edge cases (loops, openid attempts to rename, etc). And then I need to integrate it into the shop. And after that, I need to build a frontend so people can buy it from the shop *g*

At some point, I'll probably need to catch someone in chat to ask certain questions about the renaming process -- what should happen to usernames that were renamed away from, is swapping usernames one token or two, should you be able to rename a comm to a journal and vice versa, blah blah blah, buuut that can all wait a bit, because I'm sure I'll end up with still yet more questions.

Just wrote a quick script which parses out colors and the elements they apply to from a CSS stylesheet, to make it easier to match them to s2 properties.

Invoke by doing:

> css_to_s2.pl path/to/stylesheet

Can take multiple stylesheets. Actually, if it could parse out the colors from multiple stylesheets, and compare somehow pick out which colors were theme-specific, that would be pretty useful in conversion. (It does nothing of the sort now; it's just slightly easier this way than wading through the rest of the positioning, etc, code)

Mmm, not posting anywhere (else) until I figure out a better way to lay out the data. But I figure I could eventually put up a webpage as an interface in case anyone wants to use it quickly for conversion. Would need to be more careful about the input/output then, but that can wait until later.

Really simple script for you guys today: Support: Summary in Title.

That appends the the request summary into the window title which lets you the request summary show up in your browser history. Meaning, you can just type in a partially remembered request summary in your browser address bar, and it'll (hopefully) pop up the URL to the request you're thinking of.

Works on Dreamwidth and LiveJournal. (Note that it may be conflicting with some other scripts? I had to drag it to the beginning of my list in the Greasemonkey settings for it to work on LiveJournal)

I find myself utterly in love with the creativity of people who can use CSS. Because, hey, I'm happy with black on white and everything in outline form and utterly utilitarian myself, but that doesn't make for gorgeous. (My heart flutters every time I see someone take a Core 2 style, and make it really pretty, because that's what they're there for.

My heart also bounces with glee whenever someone mentions that it's easier to customize styles here than they are used to, because that's what all we did was for. Seriously, ba-dump-thump.)

Styles still not done, though. On my list, once I'm no longer focused on open-beta-blockers and the styles usability bugs being reported to us, is, in no particular order:

• sticky entry module


• custom text module


• hooks to let layout authors easily insert custom modules


• segment and arrange the wizard properly (subheaders, etc)


• pulling in information that's already available to us in other places, to be accessible using S2 (la la la la)

The one thing that I'm finding out is how easy it is to manipulate the S2 backend. Perhaps too easy. I find myself wondering whether my methods are hackish/messy and will cause problems down the line later on.

(I'm already trying to stop myself from trying to redo the grouping stuff as hashes with lots of automagic rather than arrays. Imagine being able to do property string{} module_customtext_group { grouptype= "module" }; and have it pick up the title/name/opts automatically instead of having to do a set module_customtext_group... Feel like I missed my chance on that one, now I have to do the grouping manually all the time. Bleeeeeh. (but it works, I'm happy with it))

But before that, business statistics, which I am working with, with pauamma. We've got some kind of framework hashed out, now just need to figure out how to do the selects for data collection (I say "just", but since that's the entire point of what we're doing... hah!).

The biggest problem with the SQL is that we're basically going to go through all the user accounts for data, e.g, account types. How to do this without bringing the DB to its knees? Suspect I shall need to talk to friendly neighborhood database administrator soon-ish.

Stuff like -- group by looks the easiest syntax-wise, but has to load everything in the database in one go, and that probably won't be pretty. bin/maint/stats.pl splits users into blocks, and iterates over each of the rows a block at a time. Or could you combine the two, split into blocks, and groupby within that block? Ponder, ponder, beard-stroke, etc. (Thinking out loud, will wander over to IRC and try to find a good time to talk at some point).

It's April 20. I'm so excited. Also, feeling the lack of time. (I do wish it was faster to get into the zone, though. I usually sleep Saturdays away, which leaves only part of Sunday to do code, and there's so much to do. Grr)

For instance:

    foreach var string module ( $*modules_available ) {
        # FIXME: get the values in less hacky manner
        var string section = get_plural_phrase( 0, "module_${module}_section" );
        var int order = int(get_plural_phrase( 0, "module_${module}_order" ));

        $*module_sections{$section}[$order]=[$module];
    }

(but I should probably not let something like that creep into core. Long list of if-statements, it is.)