Creating barcodes in XSL

Abstract

This document presents an XSL stylesheet to draw UPC/EAN barcodes, created by RenderX, Inc.. Similar techniques may facilitate deployment of XSL-based solutions in production environments where barcode labelling of documents is required.

Introduction

Barcodes are very simple in physical structure - just rectangular black bars separated by white spaces. Representing such a structure in XSL Formatting Objects is easy, but a really challenging problem is to calculate bar widths starting from a string of symbols to encode. However, with XSLT, it becomes plausible. To confirm this claim, a stylesheet was designed to draw UPC/EAN barcodes.

UPC/EAN (Universal Product Code / European Article Numbering) is probably the most widespread barcode system: virtually every item found in stores bears a characteristic bars-and-digits label stamped on its back. EAN codes (8 and 13 digit) are used in all European countries, while UPC ones (8 and 12 digit) are more frequent in the North America. The encoding scheme for both barcode types is the same (12-digit UPC is a subset of 13-digit EAN).

Representation of digits by bars in UPC/EAN has some peculiarities:

every digit can be represented by two alternative bar patterns;
some digits may be encoded by the alternation of bar patterns of other digits;
the last digit serves as check digit, calculated by a tricky algorithm.

An exhaustive description of the encoding algorithm can be found here.

No wonder that it took several hundred lines of XSLT code to get from a sequence of digits to the corresponding bar pattern. The result is a fairly complete XSL implementation of barcode drawing component. The stylesheet handles EAN codes (8-digit and 13-digit) and UPC codes (versions A and E). It provides for checksum calculation (including the trickiest UPC E case), and builds a complete pattern with numbers written at the bottom.

The stylesheet does more than just creating a sequence of bars: it also chooses the most traditional format for each type of code. There are subtle differences between EAN and UPC concerning digit grouping, bar lengths etc; all these are taken into account in the stylesheet.

Here is the stylesheet: barcode-EAN.xsl. Given below is a sample of its usage:

bartest.xml contains an assortment of barcodes taken from various items I found at home.
bartest.xsl is a stylesheet that converts bartest.xml to XSL Formatting Objects. It imports templates from barcode-EAN.xsl.
bartest.fo is the XSL Formatting Objects version of bartest.xml.
bartest.pdf is the result of rendering bartest.fo to PDF by XEP - a flagship product of RenderX.

You can test it yourself:

Take a bottle/can of your favourite drink out of the fridge and locate a barcode label on it;
Create a new <barcode> item in bartest.xml;
Copy all digits on the barcode label but the last one to the value attribute of a newly created item;
Save bartest.xml and apply bartest.xsl to it by means of an XSLT processor (Saxon, Xalan, XT, ...);
Convert the result to PDF by XEP. If you are not a XEP user, you can download an evaluation version here;
Look at the result and compare it with the original label. If the two patterns are not equal, send a bug report to the author.

You can also try the barcode generator online:

enter a digit string here ,
optionally - enter a check digit (the last digit in the barcode) here ,
press this button: .

Technical data

The stylesheet barcode-EAN.xsl defines a named XSLT template barcode-EAN. It takes four parameters:

value - a string of digits to encode (required). Digits may be separated by spaces, commas, dots, dashes, or colons; check digit should not be included.
code-type - one of EAN-13, EAN-8, UPC-A, or UPC-E. If omitted, the template tries to deduce the barcode type from the length of the string:
- 6-digit sequences are treated as UPC E (leading zero is prepended),
- 7-digit sequences are treated as EAN 8,
- 11-digit sequences are treated as UPC A,
- 12-digit sequences are treated as EAN 13.
module - width of the narrow bar/space; default is 0.33 mm as required by the standard.
height - pattern height (= shortest bar length); default depends on the code type.

To insert a barcode in your XSL FO document, place the following code in your stylesheet:

  <xsl:import href="barcode-EAN.xsl"/>
     ....
  <xsl:call-template name="barcode-EAN">
    <xsl:with-param name="value" select="'XXXXXXXX'"/>
  </xsl:call-template>

Here, 'XXXXXXXX' stays for the number to encode as a barcode. By specifying module and height parameters, you can control the appearance of the barcode pattern. Please refer to comments in the body of the stylesheet for more information.

(Note the single quotes around the parameter value: value type should be 'string'. If you insert an explicit value, you need an extra pair of quotes to protect this value from being treated as number; otherwise, leading zeros may result stripped).

Each barcode type expects the value parameter to contain a predefined number of digits:

EAN-13 - the data must be 12 characters long;
EAN-8 - the data must be 7 characters long;
UPC-A - the data must be 11 characters long;
UPC-E - the data must be 6 or 7 characters long. If there are 6 characters, a zero is added at the beginning.

The check digit (the last digit on the barcode labels) is not included in the argument - it will be calculated by the stylesheet itself.

The result of barcode template invocation will be an XSL FO table containing a sequence of bars, and a character string written below the bar sequence. Font size and padding of the bottom text is scaled proportionally to the pattern width. The table has an attribute that disables page breaks inside it, to prevent barcode image from accidental splitting.

Remember that tables in XSL FO are block-level elements; they cannot be included directly into a line of text. However, most practical applications require the barcode to be exactly positioned with respect to paper edges; this can be achieved by wrapping the table into an absolutely positioned block container.

The stylesheet uses an old (April 21, 99) version of XSL Formatting Objects (as used by XEP). However, changes required to bring it in conformance with the latest (March 27, 2000) XSL draft are minimal: you should change the namespace declaration on the root element and rename three properties:

n-columns-spanned becomes number-columns-spanned;
n-rows-spanned becomes number-rows-spanned;
page-break-inside="avoid" becomes keep-together.within-column="always".

Conclusion

The proposed technique has obvious advantages compared to other methods of including bar codes into XSL FO:

it produces vector graphics in output media formats:
- printed output is the sharpest possible;
- image size is reduced (less than 1 Kb per image in PDF/PostScript);
it enables a direct control or every aspect of bar code pattern - stroke widths, colors, lengths - as well as the exact positioning of the pattern on the page.

Naturally enough, this technique can be applied to other barcode systems as well. Besides UPC/EAN, we at RenderX have a working example of "Interleaved 2-of-5" barcode generator in XSL, available upon request.

In my opinion, this example demonstrates the strength of the "complete XSL" - the transformation part is used to implement a complicated algorithm, while the Formatting Objects are used to obtain a high-precision visual pattern.

Please write your comments/suggestions to Nikolai Grigoriev <grig@renderx.com>.