The Big Programming Thread - Page 325

On July 20 2013 01:54 RoyGBiv_13 wrote:


Here is the git source tree for "sed", the GNU stream editor. Scanning through the code, you'll want to look for process_files function under the execute.c, this is where the magic happens. In the early stage of processing the files, it buffers only one line at a time, malloc'ing 50 bytes for processing as per "INITIAL_BUFFER_SIZE". Later when writing back to the file, it has a buffer of 8kB, defined as FREAD_BUFFER_SIZE in order to quickly scan through a file to find where it is supposed to append each change in the file it is supposed to make. Note that this is not strictly for file streams, and is meant for any generic stream (pipe, socket, or file pointer).

You can use malloc directly for a lot of very clever memory management techniques. For example, In game design, you know that the maximum buffer size you will need, and are worried most about the worst case, so malloc'ing the entire size at the beginning of the program could save you time during a critical loop later.

My personal favorite clever use of malloc I saw a customer of mine use when doing some network optimizations. They were building a crypto stream cipher, and needed a large amount of unbounded variable memory to work on (>40MB), but couldn't pay the CPU tax when the time was needed to run the cipher. Sometimes, the network requested faster than the cipher, but sometimes slower. They malloc'd a large space near the beginning of the board initialization for the initial heap, and as the cipher got to a threshold limit in the heap, it would add another page onto the heap during a brief moment the CPU was not running the cipher because it was accessing the network peripheral on the chip. They found that where they were calling malloc() from, even within the same loop, had a large performance impact, because there is only one task allowed "inside" the kernel at a time, and adding a page to the heap requires asking the kernel for the page. When the cipher was running, it was inside the kernel, as it had to access physical registers directly.

On July 20 2013 14:11 Release wrote:
Also, on phpMyAdmin,
When I use the correct user and password combination in the config.inc file,
I get: #1045 access denied for user 'root'@'localhost' (using password: yes)

but I use a blank password,
I am able to log in but I get a warning that tells me to set a password(which would lock me out of my own server)

UPDATE mysql.user SET password=PASSWORD("NEWPASSWORD") WHERE User='root';

On July 20 2013 18:23 Morfildur wrote:


It's better to use the MySQL specific user management commands.

SET PASSWORD FOR 'root'@'%' = PASSWORD("foo");

On July 20 2013 18:45 Morfildur wrote:


Yes and no. A simple UPDATE query requires a "FLUSH PRIVILEGES" afterwards, which is easy to forget and then you wonder why your new password doesn't work. It's safer to get used to the specific commands for that as well as adding/removing users and all that.

On July 21 2013 08:09 darkness wrote:
Why do people critisise OOP?

http://harmful.cat-v.org/software/OO_programming/

Most notably Edsger Dijkstra. I don't care about the others that much.

On July 22 2013 23:40 Frigo wrote:
...
----

“Sometimes, the elegant implementation is just a function. Not a method. Not a class. Not a framework. Just a function.” — John Carmack

Yes Mr. Carmack, except if it needs several input parameters that it passes to other functions, has repeatedly recalculated subproblems, has multiple results or we want to unit test its dependencies without actually calculating it. If any of these hold, you are better off with a Method Object. So that "sometimes" is pushing it a little bit.

On July 23 2013 07:33 Release wrote:
The cool thing about OOP is that you don't have to use it. Completely up to the programmer.

The best thing about OOP imo is readability. 1 line per anything in the main (although sometimes it is better to expand. In theory, you could have 1 line for the entire program)