All the PDF libraries for Perl seem a bit barbaric -- stuck in the 1980's. You have to specify PostScript points to do layout. Java has JasperReports, Ruby has Prawn, and Python has ReportLab. Is there a non-extinct library/module that will let me make a nice looking PDF in less than a week of coding? (I'm a little frustrated by PDF::API2, PDF::Table, etc.) I don't want to generate HTML and convert it. Perl is ideal for reporting, but the main report file format is not available in a usable way. What libraries do people use?
I need:
- tables
- charts (Images)
- color
- formatting (ideally automatic, not pixel by pixel)
- headers / footers
I'm slightly open to wrapping external (non-Perl) open source tools, if absolutely needed. But not really interested in a major Java server approach. For the bounty, I want a pure Perl approach, since I want to run this on a server that I can't add more than modules to. If you have a public example that works well, please point me to it.
If LaTeX is too big, perhaps one could use Inline::Python
to wrap ReportLab, that everyone seems to like so much (I haven't used it and am not too proficient at Python).
Edit 3:
Here is Edit 2, except split into a modular style, if people like it (and if it is an kind of robust) perhaps I can publish to CPAN. For now place the .pm
file in a file structure like Inline/Python/ReportLab.pm
somewhere in your @INC
(the script's own base directory is usually in @INC
).
# Inline/Python/ReportLab.pm
package Inline::Python::ReportLab;
use strict;
use warnings;
use Carp;
use Inline::Python qw/py_eval/;
our @ISA = 'Inline::Python::Object';
sub import {
py_eval('from reportlab.pdfgen.canvas import Canvas');
}
sub new {
my $class = shift;
my $filename = shift || croak "Must specify file name to contructor";
return bless(Inline::Python::Object->new('__main__', 'Canvas', $filename), $class);
}
1;
Then a script could be something like:
#!/usr/bin/env perl
use strict;
use warnings;
use Inline::Python::ReportLab;
my $c = Inline::Python::ReportLab->new('hello.pdf');
$c->drawString(100,100,"Hello World");
$c->showPage();
$c->save();
Edit 2:
While Edit 1 is still of interest, it seems (tell me if I am incorrect!) that I have figured out how to create an instance of 'Canvas' and expose its methods directly:
#!/usr/bin/env perl
use strict;
use warnings;
use Inline::Python qw/py_eval/;
py_eval('from reportlab.pdfgen.canvas import Canvas');
my $c = Inline::Python::Object->new('__main__', 'Canvas', 'hello.pdf');
$c->drawString(100,100,"Hello World");
$c->showPage();
$c->save();
Edit 2/3: This portion is left as an example of a more manual interface. I think Edits 2/3 give a better interface which leaves the heavy lifting to the original Python class without (too much) wrapping.
Edit 1: I have now exposed some of the functionality by manually hacking in the methods. This means that for every method one wants to use, a wrapper method must be added. While this is already a feasible solution, I wonder if there isn't some easier way to expose the entire python 'canvas' class, but for now this is where I am:
#!/usr/bin/env perl
use strict;
use warnings;
use Inline Python => <<END_PYTHON;
from reportlab.pdfgen import canvas
class Canvas:
def __init__(self,filename):
self.canvas = canvas.Canvas(filename)
def drawString(self,x,y,text):
self.canvas.drawString(x,y,text)
def save(self):
self.canvas.showPage()
self.canvas.save()
END_PYTHON
my $c = Canvas->new('hello.pdf');
$c->drawString(100,100,"Hello World");
$c->save();
Using Perl, generate LaTeX, perhaps using Template::Toolkit, then call the compiler, either TeXLive or MikTeX or whatever distribution you need for your OS. There is an extension called Template::LaTeX, though you probably don't need it, which manages the build process.
LaTeX has support for all the things you need. Tables get a little interesting but there are some modern table packages which ease things (I think that its called ltxtable
). For charts (do you mean diagrams) there is a sub language called TikZ
which is spectacularly powerful.
This really is a very easy workflow, especially if you want the results to be similar every time (i.e. can use a template). In fact it really is not unlike creating HTML from a template and serving it to a browser.
Another benefit of this is that the template (and prepared source) will be portable should you need to build a report in another language.
After much thought and experimentation, I ended up writing a lot of code to wrap PDF::API2. Unfortunately this was an internal project within a company so not going to be released open source, but frankly I'd recommend using a different language (Python / Ruby), perhaps passing the data through with JSON or something. My end result is efficient, but it required a lot of coding. There is a refactoring of PDF::API2 underway on CPAN but it seems to be stalled.