Is there a document somewhere that describes all the indexing record types and their structure? I know about these:

! Index object code

!! Index source code

*INDEXES list of indexed fields

0 Root of undistributed pending index transaction records

1..N More pending transactions, where N is the first field in 0

field_name**value contains pointers to previous and next nodes and lists of record ids. I'm not quite sure what the first field is for.

etc.

I'm especially interested in the possible paths an index transaction can take. For example, a rebuild, a flush, and a background index update probably have different update paths.

Suppose, for example, I have issued an index.flush from one workstation, and I want to see the progress from another one. What would I look at? The 0 record is wiped out right away (by F.DISTRIBUTOR?). These seem to be moved into field_name*N, but then what?

Is there a document somewhere that describes all the indexing record types and their structure?

Possibly the ARev 1.x manual sets. They had some pretty detailed information on many system features that were not in later revisions. A copy of 1.x manuals is pretty valuable in this respect.

I'm especially interested in the possible paths an index transaction can take. For example, a rebuild, a flush, and a background index update probably have different update paths.

Yes and no. In the end, they all fall through F.INDEXER and F.DISTRIBUTOR for the final updates.

F.DISTRIBUTOR takes the n transactions and moves them to the IndexField*n transactions, which has the same basic format as the prior. F.INDEXER then moves them to the varying nodes and updates the data and key nodes.

There is a slight difference between MAKE.INDEX, INDEX.FLUSH and INDEX.CONTROL (background indexing). The predominance is the order files and indexes are updated, and how the process can be interupted by the user.

[email protected]

Sprezzatura, Inc.

View this thread on the forum...