Code Kills - Comments

merwok: Debugging Python

merwok@netwok.org (merwok) — Sat, 27 Feb 2010 18:43:57 -0500

What gave me this idea? Your post, fourth paragraph You should perhaps update it.

Have you updated your alias now? I love all the info mine gives me (py3k warnings, other warnings, tabs and spaces).

Cheers

merwok: Amazon S3 (or, How I Solved My Image Hosting Woes)

merwok@netwok.org (merwok) — Sat, 27 Feb 2010 18:19:59 -0500

argparse will be part of the standard library, so it’s worth learning it now. Gory details in http://www.python.org/dev/peps/pep-0389/

optcomplete won’t be updated soonish because the author is busy, but he said it would be easy to do.

Cheers

merwok: SSH and HTTPS on the same port?!

merwok@netwok.org (merwok) — Sat, 27 Feb 2010 18:16:45 -0500

By the way, you should warn that the email address we give will be displayed.

merwok: SSH and HTTPS on the same port?!

merwok@netwok.org (merwok) — Sat, 27 Feb 2010 18:15:11 -0500

Hello

I stumbled upon this and thought I’d share: http://www.rutschle.net/tech/sslh.shtml

Cheers

Stephen Paul Weber: Absolute paths must die

singpolyma@singpolyma.net (Stephen Paul Weber) — Fri, 19 Feb 2010 15:35:09 -0500

The problem with this post is that everyone who already hates absolutes paths will say "YEAH!" and everyone else will just be confused by the lack of arguments/data.

Michael Labriola: FlexUnit's Test Theories: Don't Bother

labriola@digitalprimates.net (Michael Labriola) — Thu, 18 Feb 2010 11:54:05 -0500

We use them extesnively.. that could be why we wrote them too.

As Alan correctly suggested the word theory implies scientific theories. The parameterized testing we were working on is more along the lines of what you did here and what you are looking for.

Where theories are really useful is for reversible conditions over a potentially infinite number of test cases. So, for example, we have an application that must support gregorian and hijri date systems. So, we have thousands of dates that we convert from hijri to gregorian and then back to gregorian to ensure they match.

Theories are one type of test intended to help with triangulation. They are not a solution by themselves, only when applied with standard tests that confirm functionality at known points.

Regarding lack of documentation. We are a completely community driven open-source project. We are not funded in any way and everything that exists is purely community effort, so we would be happy to have people who are willing to write help document these thigns.

HTH, Mike

Ryan Kohn: How Wolfram|Alpha makes me happy

ryan@kohn.ca (Ryan Kohn) — Thu, 11 Feb 2010 16:57:52 -0500

My favourite use for Alpha is nutritional information. You can type in something like "1 tsp olive oil, 50g cucumber, 50g tomato, 3/4 cup pasta" and it will spit out a nutritional label for you, along with all sorts of information about it.

Alan Stearns: FlexUnit's Test Theories: Don't Bother

astearns@adobe.com (Alan Stearns) — Mon, 08 Feb 2010 11:47:39 -0500

I'd thought that one colleague had found a use for them, but that turns out not to be the case. So I have to say that I'm not personally aware of anyone who's used Theories for actual testing.

David Wolever: FlexUnit's Test Theories: Don't Bother

wolever-codekills@wolever.net (David Wolever) — Sat, 06 Feb 2010 12:45:45 -0500

Thanks for the comment, Alan.

If their intention is to mimic "proving a scientific theory", then yes - the "only yielding one test" behavior does make sense.

And, since I've got you here… Has Adobe (or anyone else you know of?) found this style of testing useful? Grepping through the FlexUnit source, I was only able to find one test suite (MethodTheory.as) which makes use of Theories… So have they been used elsewhere? Are there any killer use cases I haven't though of?

Alan Stearns: FlexUnit's Test Theories: Don't Bother

astearns@adobe.com (Alan Stearns) — Fri, 05 Feb 2010 19:20:10 -0500

What you're looking for are Parameterized tests (see the Parameterized class in junit). FlexUnit will eventually support this idea of creating multiple tests from a single function given a set of test data.

Theories are meant to correspond to scientific or mathematical theories - if there's a single data point that shows the theory to be false, the entire theory is invalid (hence the single test). I agree that the error reporting is useless. That needs to be fixed.

David Wolever: Debugging Python

wolever-codekills@wolever.net (David Wolever) — Wed, 03 Feb 2010 09:51:37 -0500

What gave you the idea that py was a symlink? http://bitbucket.org/wolever/dotfiles/src/tip/profile#cl-51 : alias py="python"

I like the idea of changing it to python2.6 -3 though… I've been meaning to 3k myself and that might be the kick I need.

Merwok: Debugging Python

merwok@netwok.org (Merwok) — Wed, 03 Feb 2010 01:02:40 -0500

Minor note: using a symbolic link for shortening “python” seems a bad pratice to me. You shouldn’t mess in /usr, that’s the realm of the package manager. Setting up a shell alias does the job, and allows options as well. Snippet from my .bashrc:

alias py='python2.6 -3 -b -t -d -Qnew -Wd'

Cheers

Marguerite Wolever: codekills (and wolever.net) downtime

david@wolever.net (Marguerite Wolever) — Sat, 30 Jan 2010 13:23:34 -0500

But, in the bible Lazarus lived again... Maybe you should pray for it!

James: DVCSs and Changeset Numbering

james.cook@utoronto.ca (James) — Sat, 30 Jan 2010 12:54:06 -0500

s/or/of/

James: DVCSs and Changeset Numbering

james.cook@utoronto.ca (James) — Sat, 30 Jan 2010 12:52:16 -0500

A small note, Andrey: you can have a fairly good notion of recency if you use the longest distance to any root that is an ancestor or your commit. For example, this guarantees that if commit A is a revision of commit B, then A will have a bigger number.

I also prefer git's method, but I think it would be harmless to prepend a number like this to commit hashes in some cases, and for the utilities to simply ignore or verify that number when they see it.