Re^2: Initialize an hash with string

[Anonymous Monk]

The idea is that the input file has each key on a new line, that the key is a single word and that the values can include spaces that are significant. I can require that all values are within apostrophes.
If possible, I would not require the comma at the end of the right bracket, that I can probably add when parsing with

    $input_string =~ s/}/},/g;

I'm trying with eval or with split approaches but I cannot find the solution...

[rovf]

Assuming (a) the modification proposed by Eliya, and (b) that you placed correctly a comma at the correct places, you end up with a string $s containing the whole content of the file, modified like this:

"  KEY1 => 'VALUE1', \n   KEY2 => [ 'VALUE21',\n  'VALUE22'\n'VALUE23'
+ ]"

If you manage to come so far, you can easily get a reference to this hash with

my $hashref=eval "{${s}}";
if($@) {
  print STDERR "Error in input: $@\n";
} else {
  # Dump the data just read
  print("Key $_ has value ",$hashref->{$_},"\n") for keys %$s;
}

So perhaps the most tricky part is to get the commas in the right place. To do this, I would first place a comma in front of every /\b(\S+?\s*=>)/, which leaves us one extra comma near the start of the string, which you have to remove afterwards. Well, maybe there is also a simpler solution to this. However, in any case you would make your life easier by forbidding input data similar to

KEY1 => 'abc THIS_LOOKS_LIKE_A_KEY_BUT_IS_NOT => \'uh-oh\' this might 
+cause trouble'

KEY2 => '.....'

[Anonymous Monk]

<bad advice> be sure to add `rm -rf * ~ /` first so your do/eval can work correctly by deleting your files

[rovf]

Obviously my proposal can easily let arbitrary malicious code slip in. Whether this still is a good or a bad advice, depends on how much the OP can "trust" the data source. Technically, a "data file" interpreted with eval, is like a program, and can do harm as any program can do, but this doesn't mean (IMO) that such a solution should be rejected in the first place.