Others have already given some good advice. Here's one way of putting it all together by first converting the input SGML to well-formed XML and then using XSLT to transform that to the exact format you need.
Converting your SGML to well-formed XML
The osx
tool from the OpenSP package suggested by mzjn is a good tool for this. Since your SGML markup omits end tags, you need to have a DTD from which the correct nesting of elements can be determined. If you don't have a DTD, you need to create one. For your example input, it could be as simple as this:
<!ELEMENT toplevel o o (viewed)+>
<!ELEMENT viewed - o (#PCDATA,cite)>
<!ELEMENT cite - o (yr,pno)>
<!ELEMENT yr - o (#PCDATA)>
<!ELEMENT pno - o (#PCDATA)>
<!ATTLIST pno cite CDATA #REQUIRED>
You also need to add a proper doctype declaration to the beginning of your SGML file. Assuming you have your DTD in file viewed.dtd
.
<!DOCTYPE toplevel SYSTEM "viewed.dtd" >
With this addition, you should now be able use osx
to convert the SGML to XML. (It won't be able to convert the processing instructions which start with a /
as those are not allowed in XML, and will emit a warning about them.)
osx input.sgm > input.xml
Transforming the resulting XML to your desired format
For the above case, you could use something like the following XSLT stylesheet:
<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output method="xml" indent="yes"/>
<xsl:template match="VIEWED">
<index1>
<num viewed="{normalize-space(text())}"/>
<heading>
<xsl:value-of select="normalize-space(text())"/>
</heading>
<index-refs>
<xsl:apply-templates select="CITE"/>
</index-refs>
</index1>
</xsl:template>
<xsl:template match="CITE">
<link caseno="{PNO/@CITE}"/>
</xsl:template>
</xsl:stylesheet>