Beefy Boxes and Bandwidth Generously Provided by pair Networks
Problems? Is your data what you think it is?
 
PerlMonks  

Re^2: Parsing Problems

by ftherese (Novice)
on May 13, 2019 at 17:46 UTC ( [id://1233719]=note: print w/replies, xml ) Need Help??


in reply to Re: Parsing Problems (updated)
in thread Parsing Problems

Thank you for the suggestion. I have now ruled out the thought that this is a parsing error. I now believe it has something to do with how a function is being called:

$temp = eval qq~modes::~.$ARGV[2].qq~::flex(/$line);~; It's supposed to run a subroutine from a Perl module file in a directory (lib, opendir, readdir). Has something changed with package, lib, eval, or qq~? I'm trying to figure out how distill my example.

Replies are listed 'Best First'.
Re^3: Parsing Problems
by haukex (Archbishop) on May 13, 2019 at 18:12 UTC
    Has something changed with package, lib, eval, or qq~?

    I'm not certain, but there may have been small changes to how eval handles Unicode, but evalbytes was introduced in 5.16, so before the 5.18 that you're migrating away from. Other than that, based on this little piece of code, I don't see anything that might be different between the two versions. And whether something has changed with Text::Unaccent::PurePerl, you'd have to check what version you have installed in both environments (perl -MText::Unaccent::PurePerl -le 'print $Text::Unaccent::PurePerl::VERSION').

    If you think you're having trouble with eval, then it's best if you built its argument first, stored it in a variable, used Data::Dumper or Data::Dump to show it, and also check eval for errors using the pattern eval "...; 1" or die "eval failed: $@" (see Bug in eval in pre-5.14). (Update before posting: I see haj made a similar point.)

    However, I would strongly recommend against using eval in the first place, building Perl code from strings and trying to run it can be quite brittle, and in many cases even a major security risk. If you were to show more context (SSCCE), we could most likely suggest an alternative without eval (Update: Yep!).

Re^3: Parsing Problems
by haj (Vicar) on May 13, 2019 at 18:07 UTC

    Two things stand out here:

    • $ARGV[2] might contain un-decoded UTF-8 characters in "modern" terminals. File systems with UTF-8 characters behave differently, depending on the platform (Windows/Unix).
    • (/$line) looks weird. That needs really special values for $line to produce valid Perl!

    An idea is to print the string you're evaling, and of course checking whether eval produced an error in $@:

    my $code = qq~modes::~ . $ARGV[2] . qq~::flex(/$line);~; warn $code; eval $code; warn "Eval failed: '$@'" if $@;
      Okay! Now we're getting somewhere. In the old environment, no error is reported. In the new environment: Undefined subroutine &modes::three::a called at (eval 32) line 1. ...

      Also, I mistyped because I'm on my phone it is \$line.
Re^3: Parsing Problems
by Anonymous Monk on May 14, 2019 at 14:42 UTC

    I would really, really suggest replacing the eval() with something like

    my $name_space = "modes::$ARGV[2]"; my $code = $name_space->can( 'flex' ) or die "$name_space does not implement flex()"; $temp = $code->( $line );

    This is on the hypothesis that in fat-fingering in your example you reversed the slope of a back slash.

    When obscure code fails it is really hard to debug. I infer from your eval() example that what you are trying to do is this:

    1. Run a script whose second argument is the name of a processor for your input.
    2. Process each line of the file using a subroutine named flex() in a Perl module named "modes::$ARGV[2]"
    3. This module has already been loaded.

    If these are correct, the above code should implement it in an easier-to-read-and-debug manner. The three statements compute the name of the module, get the address of the subroutine in that module (failing if it can not be found), and call the subroutine, passing it the line from the file, and storing the result in $temp.

    This is off-topic for your question, but all your Perl should have use strict; and use warnings; near the top of the file. If your script does not do this, all sorts of bugs could lurk. Of course, if you just slam these into a legacy script it may find all sorts of things -- more than you are able to fix in one sitting. But it's worth a try -- one of the warnings/errors may tell you what your problem is.

      Thank you for your reply. This is helpful as I've traced the error to how Perl is handling the modules, @INC, and referencing subroutines.

      What my code is supposed to do is apply different sets of transformations to a text file based on options specified from the command-line:

       modes.pl [rawtext] [style] [mode](one, two, etc.) [1st variation](a, a_prime, b, etc.) [2nd variation]

      "style" is one of several directories containing nine .pm files. Each .pm file has at least the subroutines "first" and "flex." The other subroutines are variations.

      Here is the code that I'm using to load the modules:

      package modes; my $lP; my $modCount = 0; BEGIN{ $lP = "$ARGV[1]"; } use lib $lP; if(opendir(LIB, $lP)){ foreach my $l (readdir(LIB)){ unless ($l !~ /^(.*)\.pm$/){ eval qq~require ~ . $ARGV[1] . qq~::~ . $1 . qq~;~; print $@; } } }

      The error that code produces (given: ./modes.pl "filename" english three a b) in Perl 5.26.1 (error does not occur in 5.18.2) is the following:

      Can't locate english/eight.pm in @INC (you may need to install the eng +lish::eight module) (@INC contains: english /home/******/perl5/lib/pe +rl5/5.26.1/x86_64-linux-gnu-thread-multi /home/*********/perl5/lib/pe +rl5/5.26.1 /home/***/perl5/lib/perl5/x86_64-linux-gnu-thread-multi /h +ome/******/perl5/lib/perl5 /etc/perl /usr/local/lib/x86_64-linux-gnu/ +perl/5.26.1 /usr/local/share/perl/5.26.1 /usr/lib/x86_64-linux-gnu/pe +rl5/5.26 /usr/share/perl5 /usr/lib/x86_64-linux-gnu/perl/5.26 /usr/sh +are/perl/5.26 /home/****/perl5/lib/perl5/5.26.0 /home/*****/perl5/lib +/perl5/5.26.0/x86_64-linux-gnu-thread-multi /usr/local/lib/site_perl +/usr/lib/x86_64-linux-gnu/perl-base) at (eval 5) line 1.

      Setting $lP to the fully qualified PWD (or simply "." since the "english" directory is located here) fixes that error, but the subroutines are still undefined:

      Undefined subroutine &modes::three::flex

      I'm not sure why this is. I will change my code based on your recommendations, but I don't understand it well enough myself to do so yet. I borrowed the module loading code from another programmer years ago simply because it did what I needed - I never fully understood it.

        Based on what you've posed here, you definitely don't need to use the eval with require BAREWORD;, require "/path/to/filename.pm"; would be much easier in your case. Also, note that despite its several caveats, a glob might be easier than opendir in this case. Here's an example, where $stylespath is an absolute pathname (see do for the rules on when @INC will be searched and when it won't):

        use File::Glob 'bsd_glob'; my $style = $ARGV[1]; die "bad style" unless $style=~/\A\w+\z/; require $_ for bsd_glob("$stylespath/$style/*.pm");
        the subroutines are still undefined: Undefined subroutine &modes::three::flex I'm not sure why this is.

        What package statement (if any) does english/eight.pm start with? If it's e.g. package eight;, then a sub flex in that file declares a subroutine whose full name is &eight::flex.

        Note that the AM post does not name one caveat of UNIVERSAL's can method, which is that it also searches the package's ancestry via @ISA for that method - if you don't set @ISA in your packages, it's fine, but if you do, then you need to be aware of this. If you want to avoid that, then this is IMO one of those rare cases where symbolic refs can help: my $package = 'eight'; my $subref = do { no strict 'refs'; \&{$package.'::flex'} };.

        Update: That you've traced it down to this issue also helps explain what you described in your root node, that you see a difference between 5.18 and 5.26: Removal of the current directory (".") from @INC may explain your issue. But my suggestions above still stand!

Log In?
Username:
Password:

What's my password?
Create A New User
Domain Nodelet?
Node Status?
node history
Node Type: note [id://1233719]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this?Last hourOther CB clients
Other Users?
Others pondering the Monastery: (5)
As of 2024-03-28 20:51 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    No recent polls found