Refactoring browsers for C and C++

[dwks]

Yes, of course, good old search-and-replace. CTRL-H, CTRL-R, etc. Actually, I have used "search and replace in selection" on occasion. But project-wide search-and-replace, for example renaming a function, tends to go wrong. And prompting you before each replace take way too long and you get bored and click "replace" on something that you shouldn't have without realizing it. Until you try to compile or run it.

my code only generates a hundred or so bugs per feature.

Did you know, while researching preprocessor macros (there are an average of 0.6 per line, I think), I found a statistic that stated that every new line of code written has between 2 and 5 bugs in it? Scary.

[dwks]

I should mention that I downloaded Elsa, and tried this program.

#include <iostream>

int main(void) {
    std::cout << "Hello, World!" << std::endl;

    return 0;
}

Elsa requires preprocessed files, so I preprocessed that and tried parsing it with Elsa. It didn't work. g++ 4.1 worked, but not g++ 4.2. Anyway, I then got Elsa to print the elaborated abstract syntax tree. It was 692,000 lines. Ouch.

Not that the abstract syntax tree is very concise. Something like this

int main() {}

would give you at least ten or maybe fifteen lines. But still.

[matsp]

Yes, of course, good old search-and-replace. CTRL-H, CTRL-R, etc. Actually, I have used "search and replace in selection" on occasion. But project-wide search-and-replace, for example renaming a function, tends to go wrong. And prompting you before each replace take way too long and you get bored and click "replace" on something that you shouldn't have without realizing it. Until you try to compile or run it.

Yup, I recognize that...

--
Mats

[CornedBee]

(there are an average of 0.6 per line, I think),

every new line of code written has between 2 and 5 bugs in it

Lies, damn lies, and statistics.

Anyway, I then got Elsa to print the elaborated abstract syntax tree. It was 692,000 lines.

Out of curiosity, how big was the preprocessed file?

[dwks]

I found the reference that says 0.6 macros per line. http://portal.acm.org/citation.cfm?id=1172980

It has been shown that macro uses can occur as frequently as 0.6 macro uses per line [4].

I can't remember where I found that 2 to 5 thing, although I think it was in a book. Nor do I have access to the preprocessed file at the moment, but I do remember that it was about 11KB.

[CornedBee]

Well, "can occur as frequently as" and "an average of" is quite different. One is the maximum per-program average they found, the other would be an average over all the code they looked at. I'll believe the former, but the latter is absurd.

[dwks]

Yes, of course you're right. I misunderstood it.

[twomers]

So if Windows has a few kazillion lines of code and as many as 2-5 errors can occur per line with an average of 0.6, how many bugs are processed per second?