Create Reports using SVG Templates

written by randall, on Aug 6, 2009 4:37:00 PM.

A job I often encounter and have approached many ways is report creation. An oversimplified description is presenting data in a format that's useful to the recipient. When looks don't matter much, this can be as easy as creating an HTML template and merging the results of a database query, but when looks do matter the best solution isn't always apparent.

For the case when looks matter, I prefer PDF documents for the final output. The hard part is getting from raw data to PDF. Being a Python programmer I've used the reportlab library to this end with success, but often it's laborious. Also, because programming is required, non-programmers aren't able to create these reports. Many tools for end-user report creation require proprietary software and/or non-standard template formats and processes.

So what I'm looking for is:

template based
produce print ready PDF files
usable by non-programmers
use standard formats

The tools I found to meet these requirements are:

python (of course)
cairo
rsvg
Inkscape (for svg templates)

The templates are created and maintained using Inkscape in its native SVG format. In the SVG file, template fields are added as {FIELD1}, {FIELD2}, etc. A database query is run, returning fieldnames that match those in the template then python, rsvg and cairo handle template substitution and conversion to PDF.

Sample Code

def processSVGTemplate(template, data):

    filename = createNewUniqueFile()
    new_svg_file = open(filename, 'wb')

    # no need to make it difficult with xml parsing
    # just do find/replace
    svg_data = open(template).read()
    for (key, value) in data.items():
        svg_data = svg_data.replace('{%s}' % key, value)
    new_svg_file.write(svg_data)
    new_svg_file.close()

    return filename


def mergeData(template_name, data, output_path=None):
    """merge `data` with `template_path`

    returns None if `output_filename` is given or a string otherwise

    """
    template_path = os.path.join(TEMPLATE_PATH, template_name)
    # if not writing to a file, just send back a string.
    using_tempfile = False
    if output_path is None:
        using_tempfile = True
        output_path = createNewUniqueFile()
    try:
        svgt = rsvg.Handle(template_path)
        props = svgt.props
        pdf_dpi = 72.0
        width = pdf_dpi / svgt.props.dpi_x * svgt.props.width
        height = pdf_dpi / svgt.props.dpi_y * svgt.props.height
        # convert to 72 ppi
        surf = cairo.PDFSurface(output_path, width, height)
        cr = cairo.Context(surf)
        # scale the context for the change in dpi
        cr.scale(pdf_dpi / svgt.props.dpi_x, pdf_dpi / svgt.props.dpi_y)
        for rset in data:
            svg_filename = processSVGTemplate(template_path, rset)
            try:
                svg = rsvg.Handle(svg_filename)
                svg.render_cairo(cr)
                cr.show_page()
            finally:
                os.unlink(svg_filename)
        surf.finish()
        if using_tempfile:
            pdf_data = open(output_path).read()
            return pdf_data
    finally:
        if using_tempfile:
            os.unlink(output_path)

Caveats and Thoughts

rsvg required installing python-gnome-desktop on Debian (45Mb), which makes no sense to me.
I didn't do any xml parsing, but instead used string substitution. At first I used xml.dom.minidom, but it seemed needlessly complex for my needs. Maybe that will change.

Randall's Blog