Word Aligned - Latest Comments

Re: Two star programming

Lee Crocker — Sun, 09 Jun 2013 18:31:25 -0000

Don't try this at home: I have a particular application where I needed to delete a structure from a singly-linked list. I already have the address of the structure, so I don't need any key compare, and I'm guaranteed that the structure is in fact in the list, and that the list is not empty. This allowed me to condense the delete into 3 lines:

pp = &head;
while (*pp != g) pp = &((*pp)->next);
*pp = g->next;

Re: The Maximum Sum contiguous subsequence problem

seba — Sun, 12 May 2013 20:07:46 -0000

Nice article,well explained. problem i have is, i need to know between which elements of the stream,the max sequence is achieved. Anyone has an idea or suggestion for that?

Re: Angle brackets hurt my eyes

Matt Burns — Mon, 06 May 2013 04:53:42 -0000

If that's ant, there's no need to check (it's mkdir -p by default). I agree with your point though. If redundant code is too scary to look at (or takes any mental effort to parse) it's harder to spot it does nothing but add noise.

Re: “Solutions”

Gareth Rees — Wed, 17 Apr 2013 18:15:28 -0000

You need one of the Emacs/Perforce integrations (there's one with a VC-like interface, and one with a Perforce-like interface).

Re: An Exploration of the Phenomenology of Software Development

Charles Tolman — Sun, 14 Apr 2013 11:19:20 -0000

Hi Thomas,

Good to see you at ACCU. Also enjoyed Brian Marick's talk. Had an interesting talk with Tom Gilb too, who is a self-confessed fan of Descartes!

Actual slides for my ACCU talk are available at slideshare here:
http://www.slideshare.net/char...

Re: ACCU 2013

tag — Fri, 12 Apr 2013 05:40:41 -0000

I'd say you dispatched it in the sense of "dealt with a fire hazard". The show must go on! Excellent compering.

Re: ACCU 2013

Ewan Milne — Fri, 12 Apr 2013 04:58:54 -0000

Good to see you yesterday Tom. Note that I dispatched the cake in the sense of "took delivery of" rather than "ate the entire thing".

I'll be doing that over the weekend.

Re: An Exploration of the Phenomenology of Software Development

tag — Mon, 08 Apr 2013 12:32:26 -0000

Actually, it wasn't the title I was commenting on -- I meant the talk summary on the meetup and accu conference pages. Though now you mention it, maybe a more punchy title would help :-)

I hope you have a great time at the conference. I suspect your talk will end up being one of many highlights.

Re: An Exploration of the Phenomenology of Software Development

Charles Tolman — Tue, 02 Apr 2013 19:31:38 -0000

Thanks Thomas for the comments. You're right about the title - thanks to pressure of life I didn't quite get around to giving it a snazzier one!

Have just been reading Christopher Alexander's last book in the "Nature of Order" series : "The Luminous Ground", though more in reflective mode now I have mostly wrapped up the talk content. I found a fair few epiphanies in it which prompted me to put a post on my blog on the "Importance of being an Amateur". I also attached the slides as a link there if anyone is interested.

I like your site - nice title.

Re: Steganography made simple

Hemanth Kandukuri — Sat, 09 Mar 2013 16:40:41 -0000

how to get the source code of it

Re: Next permutation: When C++ gets it right

Ddura001 — Fri, 08 Mar 2013 22:29:23 -0000

What a read! But is it possible to run permutations and filter them but what each permutation means? Like rolling two dice, but the code reads it at, say, forty six instead of 4 and 6? Just a question as I'm looking someone to hire that can work with this concept.

Re: Two star programming

Bob Pelorson — Thu, 07 Mar 2013 10:56:50 -0000

Have you seen the Linux (or git) source code? It contains blatant undefined behavior, including violations of strict aliasing and integer overflow in *security critical* networking code.

As for what I've done: I work on llvm and Firefox. This is not really relevant Linus being a bad programmer though.

Re: More adventures in C++

Xan Gregg — Sun, 03 Mar 2013 10:21:04 -0000

Thanks for the write-up. With so many new features, it helps to have individual treatments rather than global feature tours.

One thing to add. For types like point that don't have a "natural" order, I try to define less<t> instead of operator<. Not a big difference, but I think the former supports the connotation that this ordering is artificial and only exists to satisfy the needs of ordered containers. And it prevents client code from using A < B as if it had a natural meaning.</t>

Re: Why Python programmers should learn Python

Test — Sun, 17 Feb 2013 17:55:23 -0000

I am java developer. I am scared of using python in a team after reading this.
It looks unmaintainable, slow and how will i code without intelli sense...

Re: Why Software Development isn’t Like Construction

Lel — Fri, 15 Feb 2013 10:37:37 -0000

le pedantic trivia face xD

Re: Two star programming

mikeb — Wed, 13 Feb 2013 19:00:42 -0000

This function never actually unlinks removed nodes from the list.

Re: Copy, load, redirect and tee using C++ streambufs

JohnKal — Mon, 11 Feb 2013 16:02:09 -0000

What is the problem with the code under the title "A bus error on my platform".

The fact that the code doesn't restore cout's streambuf before exiting is not a problem here.

Re: Two star programming

raja — Fri, 01 Feb 2013 02:09:38 -0000

shouldn't it be:
*curr = &entry->next;

???

Re: Python’s lesser known loop control

Coolgg — Wed, 23 Jan 2013 18:18:53 -0000

Yes you have to do edits twice now, but you can make that less error prone by abstracting away the parameters to the function calls (4096 here) as variables assigned above the loop.

If you did that, then most edits arguably would only need to happen once - where you assign the variables values. If you really wanted you could make it a little more generic by using the functools approach and reduce the likelihood of having to make the edit in 2 places.

I guess its a matter of preference, but I feel that the iter(object, sentinel) style is less optimal than the other 2 styles because its less known within both the python community and the bigger coding community. Even if you could make it popular among python programmers, a Java programmer for example, tasked to maintain python code would still find it confusing.

That said, your article did teach me about functools and iter that are missing in other languages - Thanks!

Re: Folded files and rainbow code

tag — Wed, 23 Jan 2013 17:33:34 -0000

Thanks, dfarmer, I do like it, especially points 08 and 09!

Re: Folded files and rainbow code

dfarmer — Wed, 23 Jan 2013 17:21:06 -0000

Thought you might like this post by Russ Cox on what might be summarized as "The Acme Way" http://9fans.net/archive/2008/...

Re: Python’s lesser known loop control

tag — Wed, 23 Jan 2013 17:16:58 -0000

Yes, that works. The repeated calls to fp.read() aren't ideal though.

Re: Python’s lesser known loop control

Coolgg — Wed, 23 Jan 2013 16:38:52 -0000

Let me try pasting that one more time:

data = fp.read (4096)
while data:
...{handle the chunk here}
data = fp.read (4096)

Re: Python’s lesser known loop control

Coolgg — Wed, 23 Jan 2013 16:36:38 -0000

I have zero python experience,
but wouldn’t the below be the more obvious
simpler solution?

data = fp.read (4096)

while data:

… {handle the chunk here}

data = fp.read (4096)

Re: Python’s lesser known loop control

robertfw — Sat, 19 Jan 2013 14:06:51 -0000

There is no code snippet that I can pull out that would really demonstrate this - as I said, the smaller and simpler the example, the less tangible the benefits (though I believe you still benefit from the decomposition of your problem into easily tested functions).

As for a more complex scenario, one of our applications stores complex product data in a graph-like structure. Different sections of the code need to be able to run a check against a given node, while also traversing the parents and children (i.e., if I "own" a child node, I also indirectly own its parent; another scenario, if any of a selected nodes parents OR children require my DOB, a form needs to present the DOB field.). The initial implementation was built imperatively and resulted in a fair bit of similar, but mostly duplicated graph traversal code.

I refactored the code to use a functional style, which resulted in a higher order function for the traversal code, which accepts two functions - the first is mapped against all of the parents and children and wraps the specific check to be run. The second is usually any() or all(), and is a fed a generator that yields the check function results.

The functional style gave us excellent separation of concerns, easily testable functions, lazy evaluation and short circuiting with generators & any/all, and a very DRY set of code. It was definitely a big win and made the code much easier to understand and extend.