Saturday, April 12, 2008

Todoist Sidebar - Firefox Add-On

Note: please see the latest update.

I recently switched from Google Notebook to Todoist for my to-do lists. I'd searched a while for a list manager that could do hierarchical tasks, and Todoist looks like the only one that does, at least for free. It does hierarchical projects too, in fact.

One hack I saw on their site described a way to load Todoist into Firefox's sidebar, basically by bookmarking a simplified view of the site and checking "Load this bookmark in the sidebar." Nice!

My joy soon soured though, when I noticed that adding a task has the side-effect of loading the Todoist website into the main browser window, browsing over whatever site you already have in the current tab. Not so nice.

Plus, I don't like the fact the button to open the sidebar (really just a bookmark) doesn't toggle -- to close the sidebar, you have to mouse all the way over to the sidebar's close button! Ridiculous! And though you can define a keyword for the bookmark which allows easy access through the address bar, there's no way to define a keyboard shortcut to hide the sidebar.

So, I put together a Firefox add-on which solves these problems, which I have creatively christened Todoist Sidebar.

Some perhaps non-obvious things to note:

  • Left-click the toolbar button to toggle the Todoist sidebar.
  • Middle-click the toolbar button to open Todoist in a new tab.
  • The default keyboard shortcut is Alt+Shift+2 on Windows and Linux, and Cmd+Shift+2 on the Mac.
  • The keyboard shortcut can be customized by modifying the string values
    todoist-sidebar.key and todoist-sidebar.modifiers in about:config, according to Mozilla's XUL:key specification. A Firefox restart is required for the changes to take effect. Setting either of these values to blank disables the keyboard shortcut.

All of this will probably be moot once Google continues their world conquest by rolling out their Calendar tasks implementation. But hey, it will be fun while it lasts...

Right now the add-on is hosted in the Firefox add-on sandbox.


10/15/08 update: Okay, Amir, the creator of Todoist, contacted me a bit ago all the way from Denmark to collaborate on the Todoist Sidebar, and he has updated it and now maintains it on the Todoist website! He blogged about it at http://amix.dk/blog/viewEntry/19333. Please go there to install. Woohoo!

5/15/08 update: Hm, it seems to still exhibit the naughty toggle behavior described in the comments on initial install. Restarting Firefox makes the problem go away. I'll look into this at some point. Lame!

4/28/08 update: Version 1.0.2 fixes the previously mentioned bug where any key press triggers the sidebar. Also the default key accelerator is now Alt+Shift+2 for Windows and Linux, and Cmd+Shift+2 for the Mac.

4/26/08 update: Please note that currently there is a bug with the default installation where every key press toggles the sidebar (see comments). This is due to my incorrectly implementing the default shortcut key. Also, the intended default is a bit problematic because it tends to interfere with the Tools menu accelerator.

I plan to change the default to Alt+Shift+2 (Cmd+Shift+2 on the Mac) when I get the chance. In the meantime, you can do this yourself by setting todoist-sidebar.key to "2" and todoist-sidebar.modifiers to "alt shift" for Windows/Linux or "accel shift" for the Mac in about:config.

4/24/08 update: Please note that some users are experiencing behavior where every key they type (or some subset) toggles the sidebar (see comments). This may possibly be remedied by fiddling with the todoist-sidebar.modifiers setting in about:config. Please be patient and I'll look into this shortly.

Tuesday, April 8, 2008

Move to Trash from the Mac Command Line

I was about to make a lengthy comment to a post on Life Hacker and decided to put most of it here instead. That site is seriously starting to take over my life.

A while back I wrote this shell function to be able to move files to the trash from the command line on my Mac, rather than condemning them to oblivion with no chance of return. This is in my .profile.

It's a little hairy looking but it does the job. Note that it attempts to append names with a date stamp if a file/directory with that name already exists in the trash, just as the Finder does.
function del() {
while [ -n "$1" ]; do
if [ ! -e "$1" ]; then
echo "'$1' not found; exiting"
return
fi

local file=`basename -- "$1"`

# Chop trailing '/' if there
file=${file%/}

local destination=''

if [ -e "$HOME/.Trash/$file" ]; then
# Extract file and extension
local ext=`expr "$file" : ".*\(\.[^\.]*\)$"`
local base=${file%$ext}

# Add a space between base and timestamp
test -n "$base" && base="$base "

destination="/$base`date +%H-%M-%S`_$RANDOM$ext"
fi

echo "Moving '$1' to '$HOME/.Trash$destination'"
\mv -i -- "$1" "$HOME/.Trash$destination" || return
shift
done
}

alias rm='del'
An interesting side-effect is that
rm <directory>
works as well, which is a little uncanny because ordinarily either rmdir <directory> (for an empty directory) or rm -rf is required.

2/23/10 update: Handle case where rm options are mistakenly specified.

7/4/08 update: Fixed problem when removing file names containing spaces.

Monday, April 7, 2008

Apologies

Sorry, I can't get enough of reading my own type (and corrections). I replied to SonOfPethuel's blog.

Friday, April 4, 2008

Create Firefox Multiple-Item Packages - maddon.py

I have a (growing) list of Firefox add-ons that I use all the time, so whenever I am customizing a clean machine, it's a tedious process browsing to each add-on's page and going through the installation procedure repeatedly.

I learned that the Mozilla framework supports Multiple-Item Packages -- multiple add-ons and/or themes grouped together in a single file. I've written a Python script which takes a list of add-ons and themes in any format I could think of and generates a multiple-item package for you -- maddon.py, for "multiple add-on." You can pass to it:

  • local .xpi/jar files
  • directories of .xpi/jar files
  • text files, with each line containing a .xpi/jar file
  • remote URLs
  • websites which will get scraped for .xpi/jar files

I use the text file VicsAddons.txt to generate my personal add-on favorites:
# Download Statusbar
https://addons.mozilla.org/en-US/firefox/addon/26
# Google Browser Sync
http://tools.google.com/firefox/browsersync/install.html
# CustomizeGoogle
https://addons.mozilla.org/en-US/firefox/addon/743
# Better Gmail 2
http://lifehacker.com/software/exclusive-lifehacker-download/better-gmail-2-firefox-extension-for-new-gmail-320618.php
# Better GCal
http://lifehacker.com/software/exclusive-lifehacker-download/enhance-google-calendar-with-the-better-gcal-firefox-extension-260074.php
# Better GReader
http://lifehacker.com/software/exclusive-lifehacker-download/trick-out-google-reader-with-better-greader-262020.php
# Adblock
https://addons.mozilla.org/en-US/firefox/addon/10
# dragdropupload
https://addons.mozilla.org/en-US/firefox/addon/2190
# Firebug
http://getfirebug.com/
# Google Gears
http://dl.google.com/gears/current/gears-osx-opt.xpi
# PicLens
http://www.piclens.com/site/firefox/tutorial_pl_ff.php

Running it produces the following output:
$ maddon.py VicsAddons.txt
Parsing 'VicsAddons.txt'
Parsing 'https://addons.mozilla.org/en-US/firefox/addon/26'
Scanning page...
Downloading 'https://addons.mozilla.org/en-US/firefox/downloads/file/24860/download_statusbar-0.9.6.1-fx.xpi'
Adding 'download_statusbar-0.9.6.1-fx.xpi'
Parsing 'http://tools.google.com/firefox/browsersync/install.html'
Scanning page...
Downloading 'http://dl.google.com/firefox/google-browsersync.xpi'
...

Created 'VicsAddons.xpi' with the following files:
download_statusbar-0.9.6.1-fx.xpi
google-browsersync.xpi
customizegoogle-0.71-fx+sm.xpi
bettergmail2_0.3.4.xpi
bettergcal_0.2.2.xpi
bettergreader_0.2.2.xpi
adblock-0.5.3.043-fx+fl+mz+ns.xpi
dragdropupload-1.5.25-fx.xpi
firebug1.0-current.xpi
firebug-1.1.0b12.xpi
gears-osx-opt.xpi
piclens-mac-release-1.6.2.1659.xpi
install.rdf

I can then drag VicsAddons.xpi to a Firefox window, and all add-ons and themes are installed in one fell swoop (restarting Firefox still required, unfortunately).

Note that since Firebug has both its current and beta versions on the same web page, maddon.py adds them both, which could be problematic. I could reference the exact desired .xpi version, but this loses the benefit of retrieving the latest version when run. Ah, trade-offs.

Here is the help.

usage: maddon.py [options] <file_or_directory> [...]

Create a multiple-item XPI. The <file_or_directory> argument(s) can be any of the following:

.(xpi|jar) file - added directly to bundle
URL, .webloc file
- if ends with '.(xpi|jar)', will be downloaded and add to bundle;
otherwise, the web page is scanned for links ending with '.(xpi|jar)'
directory - scanned for files satisfying any of the above requirements
text file - each line is scanned as if it were given as a parameter. Lines beginning
with '"', '#', '!', or ':' are considered comments

Friday, March 28, 2008

Find Duplicate Files - dups.py

I have a lot of images imported at different times and from different sources, and I wanted a quick way to find duplicates. Not finding a satisfactory (read: free) solution (though I admittedly didn't do a very exhaustive search), I took this opportunity to learn Python and came up with dups.py. Note that the file displays within a frame, so you might have to view frame source to get to the actual code.

Without arguments, dups.py checks the current directory, recursively:
$ dups.py
Duplicates found:
./Data/2004/05_4/015_12A.jpg
./Data/2004/2004.09.29 Grandma/015_12A.jpg
Duplicates found:
./Data/2002/19/uvs021219-008.jpg
./Data/2006/01_2/uvs040430-006.jpg
...
This has been tested on the Mac OS X and cygwin, and should also work with Python for Windows.

There are lots of nerdy options, like filtering by file size and following symbolic links. Try dups.py -h to see them all:
usage: dups.py [options] [<file_or_directory> ...]

Find duplicate files in the given path(s). Defaults to searching files recursively,
except for hidden files (beginning with "."), empty files, and symbolic links.

Options:
--version show program's version number and exit
-h, --help show this help message and exit
-v, --verbose verbose

Exclusion Options:
-f, --flat do not scan directories recursively
-g n, --greater-than=n
only scan files of size greater than n bytes
-l n, --less-than=n
only scan files of size less than n bytes

Inclusion Options:
-L, --follow-links follow symbolic links (warning: beware of infinite
loops)
-H, --hidden-files include hidden files
-z, --zero-files include empty files

Miscellaneous:
-D, --delete delete subsequent duplicates (files are scanned in
argument-list order)
-c, --create-rel-links
replace subsequent duplicates with relative links
(non-Windows only)
-C, --create-abs-links
same as "-c", but links are absolute
-s, --special-hidden
changes meaning of "hidden files" (-H) depending on
platform: cygwin - uses Windows file attributes
(warning: slow); win32 - files with names starting
with "." considered hidden

P.S. I hacked together a way to detect Windows hidden files from cygwin but it's ugly and slow.

4/6/08 update: I added the ability to delete duplicates (-D), and create relative (-c) or absolute (-C) symbolic links.

Saturday, March 1, 2008

Hacking Lite - Evading Coffee Shop Banners

This is mostly a note to myself and not intended to express approval of the behavior described ;-)

Occasionally I like to bring my laptop to a nearby coffee shop to get some work done without all of the distractions of my apartment. My favorite place has been a Tanner's Coffee Company within walking distance of my place. It's a little noisy sometimes, and the food isn't the freshest, but the drinks are decent and I seem to get a lot done whenever I'm there.

Their wireless offering injects an ad banner at the top of every page. This alone would not be prohibitively annoying since adblock successfully strips the ads, leaving only the banner, but what does tend to dampen the customer experience is that it breaks some sites, Google Reader in particular. Because of this, I started to do a little tinkering...

I figured they didn't inject all internet traffic, since I'm able to ssh without problems. Maybe they detect requests to servers at port 80? I toyed with the idea of using a local proxy server, blah blah blah...

Turns out, they actually filter on the user agent field within HTTP requests! This means that if you're using Firefox or Safari (or, I imagine, Internet Explorer), the banner will be injected; Opera, however, is ad-free. This also means that simply changing the user agent field that your browser declares in its HTTP requests sets you (ad) free as well.

In Firefox there are a number of ways to do this: install a Firefox extension, or simply add a string value to about:config named:

general.useragent.override

with a value like

Opera/9.26 (Macintosh; Intel Mac OS X; U; en)

as described here. It's probably a good idea to stick with a realistic user agent string as opposed to something arbitrary, since websites like Gmail may switch to less functional versions if they don't recognize your browser.

A quick way to determine your browser's user agent is javascript:document.write(navigator.userAgent).

The service responsible for the ads at this particular Tanner's (I think they're all independently owned) seems to be a company named AnchorFree. Chances are, this technique could work for ad-injection schemes used by other wi-fi spots.

Done and done. Back to high-quality coffee shop web surfing!

5/1/08 update: Okay, I'm dumb. A much easier way to do this is to add the filter

*.anchorfree.*

in AdBlock Plus. This solves the problem much more elegantly and doesn't run into issues with sites not supporting your supposed user agent.