Read and write files
| Three-argument open | |
| Read line by line | |
| Write and append | |
| Read it all at once | |
| Encoding layer |
Three-argument open
open takes a filehandle, a mode and a file name. Always use the three-argument form: it keeps the file name out of the mode string, so a name that starts with a pipe or a greater-than sign cannot change what open does.
The first examples on this page work on one small file, data.txt. Create it in the directory you run the script from, and save each snippet as demo.pl before you run it:
printf 'ok line\nerror: disk full\nfine\n' > data.txt
open my $fh, "<", "data.txt"
or die "cannot open data.txt: $!\n";
The four modes you need are < for reading, > for writing, >> for appending, and +< for read and write. The special variable $! holds the reason the call failed.
Read line by line
In scalar context the readline operator returns one line at a time, including its newline, and undef at the end of the file. That makes it the natural condition of a while loop.
open my $fh, "<", "data.txt" or die "open: $!\n";
while (my $line = <$fh>) {
chomp $line;
print "$line\n" if $line =~ /error/;
}
close $fh;
perl demo.pl
error: disk full
Do not write while (my $line = <$fh>) and then test if ($line) instead of the loop condition. A file whose last line is a single zero would stop the loop early, because 0 is false.
Write and append
The greater-than mode truncates the file, so every run starts from an empty file. Two greater-than signs append instead, which is what log writers need.
open my $out, ">", "report.txt" or die "write: $!\n";
print $out "total: 42\n";
close $out or die "close: $!\n";
cat report.txt
total: 42
The file holds the one line the script wrote. Check close as well: a full disk often reports the error there, after print has already returned success. The autodie pragma in lesson 30 turns all of these checks into automatic exceptions.
Read it all at once
Setting the input record separator to undef makes the next read return the whole file. This is fine for small configuration files and dangerous for gigabyte logs.
The demo measures config.ini, so create that file first:
printf '[main]\nport = 8080\n' > config.ini
open my $fh, "<", "config.ini" or die "open: $!\n";
my $all = do { local $/; <$fh> };
close $fh;
print length $all, " bytes\n";
perl demo.pl
19 bytes
The local keyword gives the change a scope, so the rest of the program keeps reading lines normally. Without local, every later read would return whole files.
Encoding layer
Perl works on characters, files hold bytes. Without a layer, a UTF-8 file arrives as a string of bytes and length counts bytes instead of characters. Put the layer in the mode string.
The demo prints the length of every line of names.txt, so create that file first:
printf 'Ada\nGrace\n' > names.txt
open my $fh, "<:encoding(UTF-8)", "names.txt"
or die "open: $!\n";
while (my $line = <$fh>) { chomp $line; print length $line, "\n"; }
perl demo.pl
3
5
Both names are plain ASCII here, so 3 and 5 are characters and bytes at once. Add an accented name and the two counts separate: without the layer José gives 5 bytes, with it 4 characters.
Add the same layer on output, >:encoding(UTF-8), or the characters are written as raw bytes and the file breaks for everyone else. Lesson 36 shows the one-line way to set this for all filehandles at once.
Next: File tests and metadata
Article author: Arthur Isaev
