Sort::Fields (3pm)

  use Sort::Fields;
  @sorted = fieldsort [3, '2n'], @lines;
  @sorted = fieldsort '\+', [-1, -3, 0], @lines;

$sort_3_2n = make_fieldsort [3, '2n'], @lines; @sorted = $sort_3_2n->(@lines);

Sort::Fields provides a general purpose technique for efficiently sorting lists of lines that contain data separated into fields.

Sort::Fields automatically imports two subroutines, \*(C`fieldsort\*(C' and \*(C`make_fieldsort\*(C', and two variants, \*(C`stable_fieldsort\*(C' and \*(C`make_stable_fieldsort\*(C'. \*(C`make_fieldsort\*(C' generates a sorting subroutine and returns a reference to it. \*(C`fieldsort\*(C' is a wrapper for the \*(C`make_fieldsort\*(C' subroutine.

The first argument to make_fieldsort is a delimiter string, which is used as a regular expression argument for a \*(C`split\*(C' operator. The delimiter string is optional. If it is not supplied, make_fieldsort splits each line using \*(C`/\s+/\*(C'.

The second argument is an array reference containing one or more field specifiers. The specifiers indicate what fields in the strings will be used to sort the data. The specifier \*(L"1\*(R" indicates the first field, \*(L"2\*(R" indicates the second, and so on. A negative specifier like \*(L"-2\*(R" means to sort on the second field in reverse (descending) order. To indicate a numeric rather than alphabetic comparison, append \*(L"n\*(R" to the specifier. A specifier of \*(L"0\*(R" means the entire string (\*(L"-0\*(R" means the entire string, in reverse order).

The order in which the specifiers appear is the order in which they will be used to sort the data. The primary key is first, the secondary key is second, and so on.

\*(C`fieldsort [1, 2], @data\*(C' is roughly equivalent to \*(C`make_fieldsort([1, 2])->(@data)\*(C'. Avoid calling fieldsort repeatedly with the same sort specifiers. If you need to use a particular sort more than once, it is more efficient to call \*(C`make_fieldsort\*(C' once and reuse the subroutine it returns.

\*(C`stable_fieldsort\*(C' and \*(C`make_stable_fieldsort\*(C' are like their \*(L"unstable\*(R" counterparts, except that the items that compare the same are maintained in their original order.

Some sample data (in array @data):

123 asd 1.22 asdd 32 ewq 2.32 asdd 43 rewq 2.12 ewet 51 erwt 34.2 ewet 23 erww 4.21 ewet 91 fdgs 3.43 ewet 123 refs 3.22 asdd 123 refs 4.32 asdd

# alpha sort on column 1 print fieldsort [1], @data;

123 asd 1.22 asdd 123 refs 3.22 asdd 123 refs 4.32 asdd 23 erww 4.21 ewet 32 ewq 2.32 asdd 43 rewq 2.12 ewet 51 erwt 34.2 ewet 91 fdgs 3.43 ewet

# numeric sort on column 1 print fieldsort ['1n'], @data;

23 erww 4.21 ewet 32 ewq 2.32 asdd 43 rewq 2.12 ewet 51 erwt 34.2 ewet 91 fdgs 3.43 ewet 123 asd 1.22 asdd 123 refs 3.22 asdd 123 refs 4.32 asdd

# reverse numeric sort on column 1 print fieldsort ['-1n'], @data;

123 asd 1.22 asdd 123 refs 3.22 asdd 123 refs 4.32 asdd 91 fdgs 3.43 ewet 51 erwt 34.2 ewet 43 rewq 2.12 ewet 32 ewq 2.32 asdd 23 erww 4.21 ewet

# alpha sort on column 2, then alpha on entire line print fieldsort [2, 0], @data;

123 asd 1.22 asdd 51 erwt 34.2 ewet 23 erww 4.21 ewet 32 ewq 2.32 asdd 91 fdgs 3.43 ewet 123 refs 3.22 asdd 123 refs 4.32 asdd 43 rewq 2.12 ewet

# alpha sort on column 4, then numeric on column 1, then reverse # numeric on column 3 print fieldsort [4, '1n', '-3n'], @data;

32 ewq 2.32 asdd 123 refs 4.32 asdd 123 refs 3.22 asdd 123 asd 1.22 asdd 23 erww 4.21 ewet 43 rewq 2.12 ewet 51 erwt 34.2 ewet 91 fdgs 3.43 ewet

# now, splitting on either literal period or whitespace # sort numeric on column 4 (fractional part of decimals) then # numeric on column 3 (whole part of decimals) print fieldsort '(?:\.|\s+)', ['4n', '3n'], @data;

51 erwt 34.2 ewet 43 rewq 2.12 ewet 23 erww 4.21 ewet 123 asd 1.22 asdd 123 refs 3.22 asdd 32 ewq 2.32 asdd 123 refs 4.32 asdd 91 fdgs 3.43 ewet

# alpha sort on column 4, then numeric on the entire line # NOTE: produces warnings under -w print fieldsort [4, '0n'], @data;

32 ewq 2.32 asdd 123 asd 1.22 asdd 123 refs 3.22 asdd 123 refs 4.32 asdd 23 erww 4.21 ewet 43 rewq 2.12 ewet 51 erwt 34.2 ewet 91 fdgs 3.43 ewet

# stable alpha sort on column 4 (maintains original relative order # among items that compare the same) print stable_fieldsort [4], @data;

123 asd 1.22 asdd 32 ewq 2.32 asdd 123 refs 3.22 asdd 123 refs 4.32 asdd 43 rewq 2.12 ewet 51 erwt 34.2 ewet 23 erww 4.21 ewet 91 fdgs 3.43 ewet

Some rudimentary tests now.

Perhaps something should be done to catch things like:

fieldsort '.', [1, 2], @lines;

'.' translates to \*(C`split /./\*(C' \*(-- probably not what you want.

Passing blank lines and/or lines containing the wrong kind of data (alphas instead of numbers) can result in copious warning messages under \*(C`-w\*(C'.

If the regexp contains memory parentheses (\*(C`(...)\*(C' rather than \*(C`(?:...)\*(C'), split will function in \*(L"delimiter retention\*(R" mode, capturing the contents of the parentheses as well as the stuff between the delimiters. I could imagine how this could be useful, but on the other hand I could also imagine how it could be confusing if encountered unexpectedly. Caveat sortor.

Not really a bug, but if you are planning to sort a large text file, consider using sort\|(1). Unless, of course, your operating system doesn't have sort\|(1).

Joseph N. Hall, [email protected]

perl\|(1).