Fazal Majid's low-intensity blog

Low light performance is one of the most important factors I consider when buying a camera. At one point I did an expensive switch from the Canon system to Nikon, when the D3 came out, for its amazing high-ISO performance (I returned to Canon when the 5DmkII came out).

On a popular forum for users of Micro Four Thirds cameras (which struggle beyond ISO 800), a poster recently questioned the rationale for high ISO performance, stating 99% of users will never shoot beyond ISO 800. I quickly looked at my statistics in Lightroom, and found over 54% of the photos I took in 2011 (to date) are at higher than ISO 800.

That begs the question: who is more representative, him or me? Flickr.com publishes statistics on popular camera models, but apparently not on other interesting EXIF metadata. I whipped up a quick and dirty Python script to sample recently uploaded photos from Flickr and collect the ISO speed from their EXIF tags, when available.

Of 3020 photos I sampled, fully 399 were shot at ISO higher than 800, or 13% (the 95% confidence interval is 12% to 14.4%). Thus significantly less than my proportion, but far higher than 1%.

Lightroom’s crop tool allows you to constrain the aspect ratio to a proportion of your choice, e.g. to 4:3, defaulting to the same aspect ratio as the original. The last 5 or so custom crop aspect ratios are saved, but a minor annoyance is you are unable to clear the list.

Python on the Mac and SQLite to the rescue: this simple script  lraspect.zip will reset them. If you use a non-default name for your Lightroom catalog, you will need to edit it. To run it, quit Lightroom and run the script. It will back up your catalog for you just in case.

Needless to say, I cannot be held liable if this script corrupts your catalog or eats your dog (who ate your homework), use at your own risk.

#!/usr/bin/python
import sys, os, sqlite3

# edit this to point to your LR3 catalog if you do not use the default location
lrcat = os.path.expanduser('~/Pictures/Lightroom/Lightroom 3 Catalog.lrcat')

os.system('cp -i "%s" "%s.bak"' % (lrcat, lrcat))
db = sqlite3.connect(lrcat)
c = db.cursor()
c.execute("""select value from Adobe_variablesTable
where name='Adobe_customCropAspects'""")
crops = c.fetchone()[0]
print 'aspect ratios:', crops
c.execute("""update Adobe_variablesTable
set value='{}'
where name='Adobe_customCropAspects'""")
db.commit()
print 'Custom crop aspect ratios reset successfully'

Two years after the Staceycide, the spot is still vacant. At a reported rent of $65,000 a month, that adds up to a cool $1.8M loss for the greedy landlords who pushed them out of business.

It seems they finally found a new tenant: a CVS pharmacy occupies the premises now.

I had to monitor a long-running process on a Solaris server tonight, but didn’t want to stay glued at a computer monitor. A neat trick:

ssh myserver.example.com "pwait 17601"; say "batch done"

You would replace 17601 with the process ID of the job you are waiting for, of course. That way, my Mac connects to the server, waits for the job to complete, then gives me an spoken alert when it is done. I can watch a movie, do chores or whatever during that time. I am sure there are equivalent commands to pwait for Linux.