The lw2xml program

The core of the Lexware processing pipeline is a single program called lw2xml. This is a command-line program that can be run on Mac, Linux or Windows, and is also available as an online CGI program here.

lw2xml is a single Gawk script with no dependencies other that Gawk, which is installed on all Linux machines, and can easily be installed on Mac and Windows.



Either clone the Lexweb repo:

$ git clone

or just copy the lw2xml file. (If the browser saves it as lw2xml.txt rename it to lw2xml). Make sure the file is executable:

$ chmod u+x lw2xml

Now, typing lw2xml (or ./lw2xml if $PATH is not set) should execute the program. If you do not have a gawk at /usr/bin/gawk, you can also run it with:

$ gawk -f lw2xml


Because the code in lw2xml uses some non-POSIX Awk features it will not run on the awk that comes pre-installed with Macs. You will need gawk. The easiest way to get gawk is via the Homebrew project. Just go to the Homebrew page, and follow the one-line install instructions. For this, and for running lw2xml you will need to open Terminal, an app in Utilities. (Type cd Desktop when you start Terminal, so that you are working with files on the Desktop.)

Once Homebrew is installed, just type:

$ brew install gawk

Then, either clone the Lexweb repo:

$ git clone

Or just copy the lw2xml file to your Desktop. (If the browser saves it as lw2xml.txt rename it to lw2xml.)

Execute the program with:

$ gawk -f lw2xml


lw2xml can be easily run using Gawk cross-compiled for Windows, and the CMD.EXE command prompt:

  • Download Gawk from Ezwinports and unzip on the Desktop.
  • Copy the lw2xml file to your Desktop. (If the browser saves it as lw2xml.txt rename it to lw2xml.)
  • In the menubar search box, type CMD.EXE and open it. This is the old DOS commandline. (You can also use Windows Powershell)
  • Type these commands (altering the version numbers if different). The latest CMD.EXE has command line TAB-completion, and history (with the UP arrow) which speeds things up. Basic commands: dir = view directory files, cd = change directory, copy, more = see file contents. (Substitute your Lexware file for MyLexwareFile.lw.)
cd Desktop
gawk-5.1.0-w32-bin\bin\gawk.exe -f lw2xml
gawk-5.1.0-w32-bin\bin\gawk.exe -f lw2xml MyLexwareFile.lw
gawk-5.1.0-w32-bin\bin\gawk.exe -f lw2xml MyLexwareFile.lw > out.html


With no arguments, the program prints its usage and exits. General usage:

lw2xml [ --index --s <start Line> --f <finish Line> --xml ] <lexware file>


  • --index if present, make the index (optional)
  • --s X begin processing the Lexware file at line X (optional)
  • --f Y stop processing the Lexware file at line Y (optional)
  • --xml output a (non-HTML) XML file (optional)

The program outputs the validation results to Standard Error and the processed output to Standard Out. To capture the output in a file:

lw2xml in.lw > out.html

To capture the validation results (in the Bash shell):

lw2xml in.lw 2> validation.txt 1> out.html

(On Windows the validation results must be copied from the CMD.EXE window.)

XHTML output structure

In designing an XML schema, there is no single correct structure: information may be coded in element names, or in attributes, data can be ‘flattened’ (minimal hierarchy), or ‘normalized’ (modeling with additional hierarchical levels). For the Lexware output as XHTML, all the structure information had to be stored as <div class="..."> attributes. Choices about hierarchical structure were made to facilitate the display of the data in a traditional dictionary format. For example, the root word is nested within a block of root attributes (rtattr), rather than immediately below the rt div.

Band to XML element list

The lw2xml code is written so that the possible XML hierarchies annotated in comments in the code can be extracted with this command:

grep -E ' +#>' lw2xml | sed -E 's/^ +#>//g' | sort

This gives a list of band classes and the hierarchy of XML div attributes in which the information is stored. This is not an XML schema but can help understand the XHTML structure. Another way to become familiar with the structure of the XHTML is to view a formatted dictionary file, and compare it with the XHTML source.

Simple XML version and XML schema

Using the --xml switch, a simpler, non-HTML XML output is created (without .file, ..par and com comments). This version can be more easily analyzed with XQuery. It can also be validated against an XML schema. The file lw.rnc contains the current valid XML schema, which should always be kept in sync with the Lexware grammar. The simple XML file can be validate using trang and jing. First generate the RELAX NG schema:

$ java -jar /path/to/trang.jar -I rnc -O rng lw.rnc lw.rng

Then validate:

$ java -jar /path/to/jing.jar lw.rng myLexwareFile.xml 

Alternatively, and easier on the eyes, open the XML file in Emacs. The nXML mode should launch automatically. C-c C-s C-f can be used to locate the RELAX NC schema (or navigate via menus, starting with `M-``). Invalid parts of the XML file are highlighted in red.

Programming notes

In-program comments

An attempt has been made make comprehensively comments inside the Awk script. This should assist with maintenance and modification.

Implementation choice

After initial attempts to write a parser in Awk (my language of greatest familiarity), I started working on a flex and bison parser and XML converter for Lexware files. This worked well, but with every change in allowable syntax, it required a lot of work to rewrite. It also introduced the need for a second XQuery converter from XML to HTML.

After 5 March, 2020, I started again from scratch with a simpler, single Awk script. I could develop it (and modify it) much more rapidly. The validation component is not a true parser, and simply catches non-allowed band label order, given the band label’s context, but works fine. The XHTML output allows data extraction and analysis via XQuery. And most recently (2020-06-11), an option to output as simpler XML has introduced the ability to validate the original Lexware file via a XML schema validation of the XML output (see above).

Concerning different root word classes

There are three band labels Jim uses for different kinds of root word: root (.rt), affix (.af), and root/affix (.ra). Initially, Cam had been trying to deduce differing rules for the syntax of each of these (e.g., an affix may only be a noun or verb affix, and that would determine the kinds of sub-entries that were allowed). However, on analyzing Jim's usage for Lower Tanana, he found:

  • There are affix word categories in .rt (e.g. .nsf, ..tfs)
  • There are lots of .rt word categories (..n, ..adv, etc) in .af
  • There are a few cases of both noun affix categories and verb affix categories under the same .af
  • Likewise for .ra

This recognition lead to a new, less specific strategy for validation: to treat .rt, .af, .ra with the same rule set: let any double-dot word category be under any single-dot band. The grammar was rewritten (2020-04-15) to reflect this change.