in reply to Re^3: adding lines at specific addresses
in thread adding lines at specific addresses

Yes, I think you're tight about making things harder than necessary... I came up with this solution. It makes for ugly spaghetti code, I guess, but it seems to work(TM). First, I need to slurp in the entire file since I want to match across newlines: 
my @lines = (<>);
my $text = join "", @lines;

[download]
Then, I test for each LABEL from LABEL O 0 to LABEL O 255. If the label exists, write it to another file, adding the new instructions. If it doesn't exist, just write the new instructions: 
if ($text =~ m/(   \(LABEL O 1\))(\n.+?)(   \(STOP\)\n)/s) {
    $labelo1 = $1;
    $labelo1instructions = $2;
print "$labelo1\n   NEWLIGINSTRUCTIONS HERE$labelo1instructions";
}
else {
print "(LABEL O 1)\n   NEWLIGINSTRUCTIONS HERE\n";
}

[download]
I tried Tie::File, but couldn't get my head around it. And everything I tried with it was incredibly slow, so I went back to the brute-force approach... Thanks for all your help!!

Replies are listed 'Best First'.
Re^5: adding lines at specific addresses
by graff (Chancellor) on Oct 11, 2005 at 12:37 UTC
    Well, on the whole, that's not so bad... But if you are creating 256 copies of that "if ... else ..." block, one for each possible numeric following "LABEL O", then you really have missed some important points about programming in general (understanding loops and variables) and about perl in particular (using regular expressions).

    For that matter, I think the regular expression you've shown is probably not what you really want -- try putting square brackets around the "\n." -- and don't forget to include $3 when you print stuff out.

    You also want to meet a new friend: $/ also known as "$INPUT_RECORD_SEPARATOR" (look for a description of it here: perlvar -- it's about a quarter of the way down). Based on this new information you've shown, it looks like the input data is structured in blocks, where each block ends with "(STOP)\n". You can tell perl to use that string and the end-of-record marker, instead of the default "\n", and simplify your code immensely: 

    open( IN, "some_file.tex" ) or die $!;
{
    local $/ = "(STOP)\n";

    my $expected_id = 1;

    while (<IN>) {   # read a whole block up to "(STOP)\n"
        if ( s/\(LABEL O $expected_id\)\n/$1  NEWSTUFF/ ) {
            print;  # all done with this block
        }
        else {
            print "(LABEL O $expected_id)\n  NEWSTUFF\n(STOP)\n";
            # add a new block
        }
        my $expected_id++;
    }
}  # closing this block drops the local value of $/

# now $/ is back to it's default value (in case you
# have to read other stuff in the normal fashion).

    [download]
    So, does it really need to be any more complicated than that?
[reply]
[d/l]
      "you really have missed some important points about programming in general (understanding loops and variables) and about perl in particular (using regular expressions)."

      I couldn't agree more, I'm just a humanities guy dabbling in perl scripting... Your code looks wonderful, and in fact, I had been thinking about something to that effect, but there are two problems that make such a loop impracticable:

      1. Some of the values are not given as LABEL O (number), but as LABEL C A or LABEL C a, so I'd need more than one loop.

      2. (probably trivial, but insurmountable to me) The numbers O XXX are octal numbers, and I couldn't figure out how to make perl increment the $i++ in octals.

      Again, thanks for your help!

[reply]
        **sigh**

        It seems we learn something new about your data every day... as opposed to seeing it described concisely at the outset -- even a humanities guy should be able to manage that (I used to be one myself).

        So if your file has: 

        LABEL O 1
  blah
STOP

LABEL C A
  blah
STOP

LABEL O 3
  blah
STOP

        [download]
        What are you supposed to do with that? Put "LABEL O 2" before "LABEL C A"? After it? Instead of it? Don't put it in at all? If you get "LABEL C A" and then "LABEL C C", are you supposed to fill in a "LABEL C B" as well? I suppose you probably have "LABEL X (hex number)" also, and you need to invert their order if the file contains the string "goober"...

        Whatever the next wrinkle may be, the answer is most likely "no, you don't need more than one loop". You just need to provide enough "if ... else ... else ..." conditions in the single "while" loop over data blocks in order to cover all the possible scenarios.

        (And of course, you need to be able to describe these extra conditions clearly and without ambiguity; if you can't state them coherently so a human can understand them, you won't be able to write code to do it, either. My best advice: document the algorithm first, then code it.)

        As for handling the octal stuff, try altering the top of the while loop like this: 

        while (<IN>) {
    my $exp_idstr = sprintf( "%o", $expected_id );
    if ( s/\(LABEL O $exp_idstr\)\n/$1  NEWSTUFF/ ) {
...

        [download]

        Update: sorry about the flame... and I wanted to add that there could be situations where a second (and maybe even third) pass over the data would simplify the process a lot -- e.g. on one pass, you handle all insertions of missing data blocks; on another pass, you make sure the data blocks are properly sorted; then maybe yet another pass (now that all blocks are present and in order) to add specific new lines of data to specific blocks.

        In this case, I would actually recommend that each stage/pass be written as a separate script: keep each script as simple, clear and reliable as possible, in order to do just one thing and do it right. Then run the scripts in succession over the data. (That's what pipeline commands are for: cat input | pass1 | pass2 | pass3 > output.)

[reply]
[d/l]
[select]

In Section Seekers of Perl Wisdom