Among many of freedesktop.org’s vital contributions to GUI computing on POSIX platforms, a divine one has been the uniform specification for automatically launching programs at initiation of a desktop session. Informally put (or theoretically at least), it means that you can add programs to startup in the appropriate settings, and they will be launched automatically at initiation of your session regardless of which desktop environment you use.

It’s been about 4 years since I migrated to Xfce for everyday work and it does come with a basic tool for managing startup entries (accessed through Application Menu > Preferences > Session and Startup). Still, if you need anything more than rudimentary addition/deletion of the entries you’re pretty much out of luck. For example, I wanted to start the applications in a particular order. I don’t know whether GNOME or KDE have the feature yet, but Xfce lacked the support for maintaining any such arrangement. The proper way to deal with the issue would have been to code the feature in the session manager myself, but since I barely get time to even read fortune-mod these days I instead had to rely on an ugly workaround. If you’re in the same sticky situation, here it is: Go to the autostart directory (usually ~/.config/autostart), open the startup script you’re concerned with, and edit the Exec line like this:

Before:
Exec=awn-autostart

After:
Exec=bash -c "sleep 3 && awn-autostart"

The ugliness arises from the fact that the new Exec line is merely launching the program with a delay of 3 seconds. Now, you can add another program with a delay of 4 and it will almost be guaranteed to start after the 3 seconds’ one across all desktop environments. I really hope that the specification itself would be revised soon to provide a uniform way of setting the startup order. But until that gets done, I can thank sleep for making my life simpler.

I have spent past 24 hours fiddling and messing with ext3grep in hope of recovering the lost data. To the point where I have dozens of papers lying around me with block numbers, inode values, diagrams and histograms spread all over them. To the point where I have a pretty good understanding of how Ext3 filesystem stores stuff — a fact for which on any other occasion, I would have been pretty thrilled about. To the point where every time I go to sleep, my dreams revolve around group descriptors and journal transactions with me recovering the data gleefully before the harsh reality of overwritten inodes gives me a rude awakening.

It’s about time I give up.

And as I look back over the failure for damage assessment, I recall:

• Videos hunted/downloaded over 5 years.
• Emails received/sent over 7 years.
• In total 58.3 gigabytes of personal data.

I pride myself as a geek. And for those 7 years I had been emphasizing to the non-geeks again and again the importance of backups and implications of data loss. The first and penultimate data loss occurred in 2002, when I accidentally formatted my hard drive. Since then, I have been paranoid about the safety of my digital life and own a terabyte of storage space specifically for this purpose. Yet, when disaster struck, it was my own hands that orchestrated it. I could only helplessly feel the goosebumps and tingling sensations in my spine as I grasped the repercussions of my actions.

Right now, as I reminisce about the 7 years, I see lying in front of my the hard disk which perhaps — in plethora of bits containing zeros and ones — still has the data beyond my reach. The deepest onset of nostalgia I have ever had, combined with the gravest sense of loss makes me loathe with a burning passion this piece of machine I am typing on. Effectively I am back to square zero of my geekdom. Effectively, I have nothing now but vague and distressful recollections of what used to be my data kingdom.

“What’s saved affords no indication of what’s lost.” — Edward Bulwer-Lytton

rm -rf ~

The scenario is somewhat messy. But it goes something like: I forgot to take backup of a particular folder on my portable drive along with the other backups, reformatted my local drive few times for new Fedora, installed it and popped my portable hard disk back in for the harrowing realization that almost 60 gigabytes of personal data; including mails, chats, pictures, videos and downloads is dead and gone.

There is a reason why I never evangelize Fedora much: it’s far from perfect. There is also a reason why I use Fedora for my everyday Linux-ing: it’s good when it starts working. The two seemingly conflicting viewpoints are not mutually exclusive, since Fedora’s instability as well as its appeal can be termed in one word: bleeding-edge.

The biannual cycle of backup-reinstall-reconfigure for each new Fedora release works fine for me. It takes a day or two to get everything back to the way want them to be, but the trade-off is almost always worth it. Consider this, upgrading from a Fedora version to the next will almost never work seamlessly. Nevertheless, the re-installation mantra allowed me to use PulseAudio’s “perfect” setup much before most of the other distros’ users. The fact that I got a clean start for the new sound architecture also explains why I never had any incompatibility issues with it and consequently, my status as one of its huge fans.

Another reason why I’m still hanging on to Fedora is because of the familiarity factor, as I no longer have the ample time for trying out newer distributions just for the sake of it. Linux From Scratch was a mighty fun experience, but keeping it up-to-date was nothing short of a Herculean task. Gentoo sounds very appealing, but I don’t want to download a DVD, go through the hassle of installing and getting familiar with the new distro and then emerge a whole universe of updates on an internet connection as reliable as Fedora on Hurd. The perfect solution would be to install a new release of Gentoo but unfortunately, it’s been over an year now since I started waiting for one.

Which brings us to Fedora 11 — lovingly nicknamed after everyone’s favorite Spartan king.

The predominant aspects of the new release which affect me as a user are the Ext4 filesystem and the Kernel Modesetting feature. So far, I am not even using the former. Yes, yes, I know I bragged about bleeding-edge before, but let me explain why I did not go for it in this particular case:

• Filesystem is the most critical aspect of a system for me. I can toy around and experiment with everything but this holy grail. If my audio fails, I can try fixing it. If X stops working, ditto. If my data is corrupted, I’m FUBAR.
• Anaconda crashed when I tried to update the boot-loader configuration for a system with Ext4 partitions.
• I couldn’t find a way to use Ext4 without using the Logical Volume Manager. I like my partitions as /dev/sdaX and entries such as /dev/mapper/yourmom/blahblahblah/finallythedrive in the fstab file turn me off.

Nevertheless, KMS made up for the Ext4 — or lack thereof.

• Booting is now prettier.
• Working in runlevel 3 is so much more efficient because of all the extra space provided by higher resolution.
• Switching virtual terminals and X sessions is a breeze.
• Suspend/Resume is a bit more stable.

Other significant changes include Xfce 4.6, Firefox 3.5 and the reworked volume management in Pulse. Overall, I’m pretty satisfied with my decision to upgrade, and here’s to the hoping that the next 6 months will go as smooth as the previous ones.

A couple of days back, I had to buy a laptop for my dad. Now, deciding one for him was infinitely easier than doing so for myself since the obsession with smaller screens wasn’t playing any role here. In fact, what he ended up liking was a 15.6″ behemoth which, upon comparison, could easily swallow my 12.1″ and 8″ notebooks and still have space left for my cellphone.

The purchase was made at eXtra. Before my dad finalized it though, the salesman asked me if I would like to have the laptop setup with “original Windows Vista, original Anti-Virus software with all the updates, original office and configured with all the drivers for SR 365 only”.

This immediately raised a few points in my head:

• Dad happens to be as much of a computer guy as I am an electronics’ (although I have some extra credentials, such as scoring a hat-trick of flunking performances at the university in a single course related to electronics). I really didn’t want him to be bothered with all the “Windows Genuine Advantage” pile of crap later on.
• I would be saved the trouble of downloading, installing, cracking, patching and updating a “clean” Windows installation — regardless of the fact that I find it quite amusing whenever someone refers to a Windows installation as “clean”.
• Around $100 would be a good bargain considering retail prices of all the softwares mentioned.

Then, a few counter-points:

• Until that moment, I had been totally oblivious to Compaq laptops coming without having any pre-installed operating system. In fact, that’s one of the reasons why I myself had settled for an OS-less Fujitsu-Siemens notebook few years ago, which gave me very handsome physical as well as technical specs for the cash I spent. Back then, I was also pleasantly surprised when my Linux From Scratch system got migrated to the new machine using only bash, netcat and tar; making it usable on the very day of notebook’s acquisition.
• I harbor a particular distrust for salesmen who speak too fast.

The counter-points outweighed the originals, and I decided to go with an empty laptop. Got home, downloaded and burned a cracked copy of XP SP3, only to find out that the laptop wasn’t empty at all. It already had a working Vista & Co. on it which only required setting up the initial time and localization settings. Immediately, I recalled other unsuspecting customers at the counter who did pay the extra charges for getting their laptops “ready”.

Fortunately for my dad, my time-telling prowess wasn’t as valuable as the salesman’s so I didn’t ask him $100 for it. Unfortunately, he’ll now have to cope with Vista.

“Windows: Microsoft’s tax on computing neophytes.”

Changelog:

• As can be seen from the screenshot above, graphs can be turned on through the options menu. Implementing them resulted in two useful modules:
  • scrolledcanvas: Provides a derived Canvas class which has built-in support for scrolling oversizes images.
  • roundedrectangle: Provides functions for drawing rectangles with rounded corners on a Canvas or Image. Optionally, text can be given which will be prettily centered (and truncated upon requirement) in the drawn shapes.
• More minor code enhancements and bugfixes.

You can find both the modules and the application itself here. All code is released under the PSF license so feel free to use it any way you want. Oh, and if you still haven’t figured out why anyone would be interested in the SMS stats in the first place, here’s a little quote for you:

“Statistics are like a bikini. What they reveal is suggestive, but what they conceal is vital.”

Iterable classes are one of the features which make Python code more readable. Simply put, they let you iterate over a container a la:

1
2

for s in ("Spam", "Eggs"):
	print s

Here, s iterates over the tuple printing the words one by one.:

Spam
Eggs

Now comes the interesting part: How do I make my own classes iterable? The official Python Tutorial gives a working example for how to do it:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

class Reverse:
	"Iterator for looping over a sequence backwards"
	def __init__(self, data):
		self.data = data
		self.index = len(data)
 
	def __iter__(self):
		return self
 
	def next(self):
		if self.index == 0:
			raise StopIteration
		self.index = self.index - 1
		return self.data[self.index]
 
value = Reverse('spam')
for char in value:
		print char

Output:

m
a
p
s

The example appeared perfectly fine to a beginner like me. However, since I’m just kinda twisted in the head, I added a new line in the for loop:

16
17
18
19

value = Reverse('spam')
for char in value:
	if char in value:
		print char

Which resulted in the (quite unexpected) output:

That’s it. Nothing. Even though the code should make perfect sense and does work in case of built-in types. For example:

1
2
3
4

tup = ("Spam", "Eggs")
for s in tup:
	if s in tup: 
		print s

So daisy-ly gives:

Spam
Eggs

The culprit in case of tutorial’s example for user-defined iterators? After toying around the code sample a little, here’s what I pinned down:

• On the nested lines where another iterator is required, the Reverse class is supposed to return instance of an iterator which would define the next() method for returning successive values.
• Since the Reverse class returns only itself in this scenario, the self.index variable is shared among iterators of the Reverse('spam').
• As a result, Reverse.next() raises the StopIteration in the nested condition.

Once I understood the underlying problem, some further head-scratching and a can of malted drink resulted in the solutions:

• Return a copy for the iterative functions instead of the instance itself:

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

  import copy   class Reverse: "Iterator for looping over a sequence backwards" def __init__(self, data): self.data = data self.index = len(data)   def __iter__(self): return copy.copy(self)   def next(self): if self.index == 0: raise StopIteration self.index = self.index - 1 return self.data[self.index]   value = Reverse('spam') for char in value: if char in value: print char

  Pro: Less strain on the programmer, only a couple of extra lines of code are needed.
  Con: copying the instance can be expensive in case of larger containers.

• Use another class:

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

  class Reverse: "Iterator for looping over a sequence backwards" def __init__(self, data): self.data = data   def __iter__(self): return ReverseIter(self)   class ReverseIter: def __init__(self, inst): self.inst = inst self.index = len(self.inst.data)   def next(self): if self.index == 0: raise StopIteration self.index = self.index - 1 return self.inst.data[self.index]   value = Reverse('spam') for char in value: if char in value: print char

  Pro: Since the whole container is not copied, only the index is unique among iterators — less burden on the memory.
  Con: Not everyone likes defining new classes.

Both solutions worked equally well and resulted in the same output (the expected one this time):

m
a
p
s

The choice of either solution is solely dependent on the programmer’s preference. As a side note, after equating Python programming with carnal activities in few of my previous posts, I’m gonna take it to the next step and finally tag this post accordingly.

Update: New version

Improvements in the new version:

• Previous version hung up while calculating the statistics. The new version dispatches a thread for the dirty work and keeps the user interface responsive with a “Processing” notification.
• Contact stats are sorted in descending order by the number of messages per each contact.
• Code improvements for making it more “Pythonic”.

inboxstats.py

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86

# -*- coding: utf-8 -*-
"""Script for printing trivial statistics about inbox, such as:
	Number of texts
	Number of unique contacts who sent the texts
	Number of texts sent by respective contacts
"""
 
__author__ = "Kamran Riaz Khan"
__email__ = "krkhan@inspirated.com"
__version__ = "0.2"
__copyright__ = "Copyright (c) 2009 Kamran Riaz Khan"
__license__ = "Python"
__status__ = "Production"
 
import appuifw, e32, inbox, thread
 
def exit_key_handler():
	"Release app_lock."
	app_lock.signal()
 
def parse_inbox_stats(stats):
	"""Parse the inbox statistics,
	Updates the stats dictionary with:
		sms-count : Number of texts
		sms-contacts: List of tuples with following pairs:
			Name of contact, Number of corresponding
			(ordered according to decreasing number of texts)"""
	curr_inbox = inbox.Inbox()
	messages = curr_inbox.sms_messages()
	contacts = {}
 
	for i in messages:
		address = curr_inbox.address(i)
		if contacts.has_key(address):
			contacts[address] = contacts[address] + 1
		else:
			contacts[address] = 1
 
	contacts = contacts.items()
	contacts.sort(lambda x, y: cmp(x[1], y[1]))
	contacts.reverse()
 
	stats["sms-count"] = len(messages)
	stats["sms-contacts"] = contacts
 
def print_inbox_stats(content, stats):
	"""Print inbox stats in the content Text field,
	Remembers the cursor position of Text before the call
	and points at it again after updating the content."""
	pos = content.get_pos()
 
	statsmap = [
		(u"SMS Count", unicode(stats["sms-count"])),
		(u"Unique Contacts", unicode(len(stats["sms-contacts"]))),
		(u"", u"")
		]
 
	statsmap += [(k, unicode(v)) for k, v in stats["sms-contacts"]]
 
	for i in statsmap:
		content.style = appuifw.STYLE_BOLD
		content.add(i[0] + (i[0] and u": " or u""))
		content.style = 0
		content.add(i[1] + u"n")
 
	content.set_pos(pos)
 
if __name__ == "__main__":
	content = appuifw.Text()
	appuifw.app.title = u'Inbox Stats'
	appuifw.app.body = content
	appuifw.app.exit_key_handler = exit_key_handler
 
	stats = {}
	t = thread.start_new_thread(parse_inbox_stats, (stats,))
 
	content.style = appuifw.STYLE_ITALIC
	content.add(u"Processing text messages...n")
	thread.ao_waittid(t)
	content.add(u"Done!nn")
	content.style = 0
 
	print_inbox_stats(content, stats)
 
	app_lock = e32.Ao_lock()
	app_lock.wait()

Update: New version

Self-indulgence is what I do best. It usually results in me trying to figure out random statistics about my personal life; e.g., graphs about which hours of day I’m mostly awake on and pie-charts about my bathroom habits. Such stuff doesn’t only make me feel more important than I actually am, but also polishes my fundamental math skills which were lost while trying to calculate average number of viruses a Windows user is hit by on an yearly basis.

Texting is what I do second best. Combine the two of my most productive practices and the need emerges of having a way to produce useless statistics about my cell phone’s inbox. This is where PyS60 comes to the rescue. In my previous post I praised Python’s sex appeal. Here’s the demonstration:

• Total time spent with Python: Less than a week
• Total time spent with PyS60: Less than a minute
• Total time spent with Symbian development: Less than never

And still, even a total n00b like me could easily accomplish what he wanted to, using only the library reference manual and 70 lines of understandable code:

inboxstats.py

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

"""Script for printing trivial statistics about inbox, such as:
	Number of texts
	Number of unique contacts who sent the texts
	Number of texts sent by respective contacts
"""
 
__author__ = "Kamran Riaz Khan <krkhan@inspirated.com>"
__version__ = "$Revision: 0.1 $"
__date__ = "$Date: 2009/05/10 15:30:00 $"
__copyright__ = "Copyright (c) 2009 Kamran Riaz Khan"
__license__ = "Python"
 
import appuifw
import e32
import inbox
 
def exit_key_handler():
	"Release app_lock."
	app_lock.signal()
 
def inbox_stats():
	"""Parse the inbox statistics,
	Returns the dictionary:
		sms-count : Number of texts
		sms-contacts: Dictionary with the pairs:
			contact-name : Number of texts from contact"""
	cur_inbox = inbox.Inbox()
	messages = cur_inbox.sms_messages()
	contacts = {}
 
	for i in messages:
		address = cur_inbox.address(i)
		if contacts.has_key(address):
			contacts[address] = contacts[address] + 1
		else:
			contacts[address] = 1
 
	return {
		"sms-count" :  len(messages),
		"sms-contacts" :  contacts
		}
 
if __name__ == "__main__":
	content = appuifw.Text()
	appuifw.app.title = u'Inbox Stats'
	appuifw.app.body = content
	appuifw.app.exit_key_handler = exit_key_handler
 
	stats = inbox_stats()
	statsmap = (
		(u"SMS Count", unicode(stats["sms-count"])),
		(u"Unique Contacts", unicode(len(stats["sms-contacts"]))),
		)
 
	for i in statsmap:
		content.style = appuifw.STYLE_BOLD
		content.add(i[0] + u": ")
		content.style = 0
		content.add(i[1] + u"n")
 
	content.add(u"n")
	for k, v in stats["sms-contacts"].iteritems():
		content.style = appuifw.STYLE_BOLD
		content.add(k + u": ")
		content.style = 0
		content.add(unicode(v) + u"n")
 
	app_lock = e32.Ao_lock()
	app_lock.wait()

Which gives me:

Remember the good old days when playing Snakes on mobile used to be about the most productive thing you could do in a classroom? Well, those days are back, but this time taking guise of another fun reptilian phenomenon: Python for Series 60. If you need to develop/prototype applications on Series 60 devices while having some real fun, you might find PyS60 to be the best thing Nokia did since 1100.

In past, I have tried demystifying the beast known as Symbian development. Truth be told, I really ended up wishing that I had never attempted to do so in the first place. The whole development process was:

• Extremely bloated: You need about a supercomputer to crunch out one innocent little SIS file without waiting for eons.
• Error prone: Put the SDKs in a non-standard path and you’re foobar-ed.
• Intimidatingly cryptic: For a beginner (and by beginner I mean beginner to Symbian, C++ experience apparently proves to be of no help over here), reading Symbian C++ is more or less like reading Perl. Especially since the humor that has developed over the decades resembling Perl code to line noise doesn’t lose any of its appeal here either.

For example, to create a simple notification which would read “Spam and eggs”, I would need to spend about 8 hours downloading, configuring, compiling, comprehending, troubleshooting the development tools. Further 4 for trying to understand how to accomplish something so simple in Symbian code. Granted, such painful development procedures might be required in some scenarios (e.g., where speed is a factor or where masochistic programmers prevail); in PyS60, producing the notification was as simple as:

import appuifw
appuifw.note(u"Spam and eggs", "info")

A mammoth two lines of code which I can easily understand without even referring to a book — I feel so cheap.