Perl Style Guidelines

Use Test::More for test scripts while using Test::Count annotations

One should use Test::More for new test scripts, while using Test::Count ( http://beta.metacpan.org/module/Test::Count ) "# TEST" annotations. Some of the old test scripts under t/*.t are still using Test.pm, but it should not be used for new code.

Any bug fixes or feature addition patches should be accompanied with a test script to test the code.

Avoid trailing statement modifiers

One should not use trailing "if"s "while"s "until"s, etc.

Bad:

print "Hello\n" if $cond;

Good:

if ($cond)
{
    print "Hello\n";
}

Avoid until and unless

"until" and "unless" should be spelled using "if !" or "while !" or alternatively "if not" or "while not".

Make sure you update the "MANIFEST" file with any new source files

All the new source files should be places in the "MANIFEST" file in the core distribution. Note that I am considering to make use of "MANIFEST.SKIP" instead, which would not necessitate that in general.

Make sure to update the "Changes" (or equivalently named) file

A patch should also patch the "Changes" file (whose name may vary) with the explanation of the change. A Changes file should not be automatically generated. Note that due to historical reasons, the exact format of the Changes varies between different projects of mine and you should try to emulate the style and format of the one of the CPAN distribution in question.

Test programs should not connect to Internet resources

As a general rule, test programs should not connect to Internet resources (such as global web-sites) using LWP or WWW::Mechanize or whatever, and should rely only on local resources. The reasons for that are that relying on such Internet resources:

  • May fail if the machine does not have a fully open Internet connection.

  • Will add load to the hosts in question.

  • Such Internet resources can fluctuate in their content and behaviour, which may break the tests.

Bug fixes and new features should be accompanied by new automated tests

One should add new assertions to t/*.t to provide regression tests for bug fixes and new features. These tests should fail before the core code is modified and succeed after it. Needless to say, the previously existing test assertions should all pass as well.

For more information, see:

Other elements to avoid

C Style Guidelines

Here are some style guidelines for new code to be accepted into XML-LibXML:

4 Spaces for Indentation

The source code should be kept free of horizontal tabs (\t, HT, \x09) and use spaces alone. Furthermore, there should be a 4 wide space indentation inside blocks:

if (COND())
{
    int i;

    printf("%s\n", "COND() is successful!");

    for (i=0 ; i < 10 ; i++)
    {
        ...
    }
}

Curly Braces Alignment

The opening curly brace of an if-statement or a for-statement should be placed below the statement on the same level as the other line, and the inner block indented by 4 spaces. A good example can be found in the previous section. Here are some bad examples:

if ( COND() ) {
    /* Bad because the opening brace is on the same line.
}
if ( COND() )
    {
    /* Bad because the left and right braces are indented along with
    the block. */
    printf(....)
    }
/* GNU Style - fear and loathing. */
if ( COND() )
  {
    printf(....)
  }

Comments should precede the lines performing the action

Comments should come one line before the line that they explain:

/* Check if it can be moved to something on the same stack */
for(dc=0;dc<c-1;dc++)
{
    .
    .
    .
}

TODO: Fill in

One line clauses should be avoided

One should avoid one-line clauses inside the clauses of if, else, elsif, while, etc. Instead one should wrap the single statements inside blocks. This is to avoid common errors with extraneous semicolons:

/* Bad: */
if (COND())
    printf ("%s\n", "Success!");

/* Good: */
if (COND())
{
    printf ("%s\n", "Success!");
}

/* Bad: */
while (COND())
    printf("%s\n", "I'm still running.");

/* Good: */
while (COND())
{
    printf("%s\n", "I'm still running.");
}

Identifier Naming Conventions

Here are some naming conventions for identifiers:

  1. Please do not use capital letters (including not CamelCase) - use all lowercase letters with words separated by underscores. Remember, C is case sensitive.

  2. Note, however, that comments should be phrased in proper English, with proper Capitalization and distinction between uppercase and lowercase letters. So should the rest of the internal and external documentation.

  3. Some commonly used abbreviations:

max - maximum
num - numbers
dest - destination
src - source
ptr - pointer
val - value
iter - iterator
idx - index
i, j - indexes

Don’t comment-out - use #if 0 to temporarily remove code

Code should not be commented-out using gigantic /* … */ comments. Instead, it should be out-blocked using #if 0…#endif.

In Perl code, one can use the following POD paradigm to remove a block of code:

=begin Removed

Removed code here.

=end Removed

=cut

No declarations after statements

One should make sure there are no declarations after statements in the ANSI C code. If you’re using gcc, you can make sure this is the case by adding the flags "-Wdeclaration-after-statement -Werror" to "CCFLAGS" in the makefile.

Bad:

int my_func(int arg)
{
    int var;

    printf("%s\n", "Foo");

    /* Declaration after statement. */
    int other_var = var;

    return;
}

Better:

int my_func(int arg)
{
    int var;
    int other_var;

    printf("%s\n", "Foo");

    other_var = var;

    return;
}

Comments should have an empty space after the comment leader

Comments in Perl, C, Python, Ruby, and other languages should have an empty space after the comment leader.

Bad:

#Print a value.
print "Hello\n";
/*Print a value.*/
printf("%s\n", "Hello");

Better:

# Print a value.
print "Hello\n";
/* Print a value. */
printf("%s\n", "Hello");

sizeof(var) is preferable to sizeof(mytype_t)

Given the choice between sizeof(var) as well as sizeof(*var) and sizeof(mytype_t) where mytype_t is a type, the former should be preferable. This way, if the type of the variable changes, one does not need to fix the sizeof(…).

sizeof() must always be enclosed in parentheses

Do not write sizeof int, sizeof mystruct_t etc. Instead write sizeof(int), sizeof(mystruct_t) .

Types should end in “_t” ; Raw struct definitions in “_struct”

New typedefs should call their types in names that end with a “_t”:

typedef int myint_t;
typedef struct
{
    .
    .
    .
} mystruct_t

Prefer doing typedef struct { … } mystruct_t to declaring a struct separately. If a struct’s declaration is still needed (e.g: if it contains a pointer to itself) it should:

  1. Have an identifier that ends with “_struct”.

  2. Be typedefed into a type (that ends with “_t”): typedef struct my_struct_struct my_struct_t;.