4. Processing Phases

4.1 The Startup Phase

After the system is started, xindy reads in the index style that is passed as a command line argument. Each require command is executed and the internal data structures representing the index style are built up. The index style must completely be read in before the raw index can be read.

4.2 The Processing Phase

The processing phase starts with reading the complete raw index. The name of the raw index file must be passed via the command line. All index entries are read in and pre-processed. All attributes and cross reference classes are checked if they are already known to the system. All strings representing location references are matched against all known location classes. Appropriate warnings are issued, if errors are encountered.

After the raw index has completely been read in, the location references of each index entry are merged, separated and sorted and the building of ranges takes place. This phase is the most complex one and we will describe it in detail.

1. All location references are separated according to the class they belong to. These groups are called location class groups. Possible groups are all defined location and crossref classes. See the commands define-location-class and define-crossref-class for a description how these classes can be defined. The classes are sorted according to an order that can be defined with the command define-location-class-order.
2. The further processing of each location class group is different for the location classes and the crossref classes.
   • Cross references are sorted lexicographically based on the ISO-Latin alphabet.
   • To illustrate the processing of location references we assume the following list:

     13, 14, 15, 18, 12, 13, 14, 16, 14, 16

     The location references in italics own the attribute `important, those with in boldface have attribute `definition', and all others are own the attribute `default'. Imagine, the attribute groups were defined with the commands

        (define-attribute-groups (("definition" "important")
                                  ("default")))
     
        (merge-to "definition" "default" :drop)
        
     

     See commands define-attributes and merge-to for a detailed description. The substitution rules are applied. This means that location references 13 and 14 with attribute `important' are substituted by the location references 13 and 14 with attribute `definition'. Substitution means removing from the list of location references. Substitution occurs because the definition of the attribute groups implicitly defines "definition" substitutes "important". The resulting list is now

     13, 14, 15, 18, 12, 16, 14, 16

     The merge-to rules are applied. Their meaning is to make location references appear with another attribute as well, but only in the function of supporting the building of ranges. They disappear after the ranges are built. The location references that cause new location refererences to be added are called parents, whereas the new ones are called childs. The example rule results in the adding of all refernces with attribute `definition' to the attribute `default' which results in the list

     13, 14, 15, 18, 12, 16, (13), 14, (15), 16, (18)

     The childs are put in parenthesis since they may only be used to build up ranges. For each attribute we now try to build ranges. Since the switch :drop was specified we must start with the attribute `default', because a successful merging of location references may result in dropping the parents. This results in the range `13--16'. The childs (13) and (15) were used in the building of ranges, so their parents 13 and 15 have to be removed from the list of location references. This step would be omitted if the switch :drop were not specified. After unsucessfully trying to build more ranges and dropping the location references 13, 15 and (18)--which was only meant to build ranges--we obtain the list

     14, 18, 12, 16, 13--16

     Finally the attributes are brought into the right order. In our example the location references of the first attribute group are merged and sorted lexicographically resulting in two attribute groups

     (12, 14, 16, 18) (13--16)

After all index entries have been processed the letter groups are formed and the index entries and location references are transformed into tree like structures as defined in the index style.

4.3 The Markup Phase

After the index has completely been processed, the markup phase traverses the tree-like structure of the index. Each step triggers the appropriate markup events resulting in the emitting of markup tags. This phase can be traced by using the command line option -t.