man stag-diff (1): finds the difference between two stag files

SYNOPSIS

stag-diff -ignore foo-id -ignore bar-id file1.xml file2.xml

DESCRIPTION

Compares two data trees and reports whether they match. If they do not match, the mismatch is reported.

ARGUMENTS

-help|h

shows this document

-ignore|i ELEMENT

these nodes are ignored for the purposes of comparison. Note that attributes are treated as elements, prefixed by the containing element id. For example, if you have

  <foo ID="wibble">

And you wish to ignore the ID attribute, then you would use the switch

  -ignore foo-ID

You can specify multiple elements to ignore like this

  -i foo -i bar -i baz

You can also specify paths

  -i foo/bar/bar-id

-parser|p FORMAT

which parser to use. The default is XML. This can also be autodetected by the file suffix. Other alternatives are sxpr and itext. See Data::Stag for details.

-report|r ELEMENT

report mismatches as they occur on each element of type ELEMENT

multiple elements can be specified

-verbose|v

used in conjunction with the -report switch

shows the tree of the mismatching element

OUTPUT

If a mismatch is reported, a report is generated displaying the subpart of the tree that could not be matched. This will look like this:

REASON: no_matching_node: annotation
  no_matching_node: feature_set
    no_matching_node: feature_span
      no_matching_node: evidence
        no_matching_node: evidence-id
          data_mismatch(:15077290 ne :15077291): evidence-id AND evidence-id

Due to the nature of tree matching, it can be difficult to specify exactly how trees do not match. To investigate this, you may need to use the -r and -v options. For the above output, I would recommend using

  stag-diff -r feature_span -v

ALGORITHM

Both trees are recursively traversed... see the actual code for how this works

The order of elements is not important; eg

  <foo>
    <bar>
      <baz>1</baz>
    </bar>
    <bar>
      <baz>2</baz>
    </bar>
  </foo>

matches

  <foo>
    <bar>
      <baz>2</baz>
    </bar>
    <bar>
      <baz>1</baz>
    </bar>
  </foo>

The recursive nature of this algorithm means that certain tree comparisons will explode wrt time and memory. I think this will only happen with very deep trees where nodes high up in the tree can only be differentiated by nodes low down in the tree.

Both trees are loaded into memory to begin with, so it may thrash with very large documents

AUTHOR