Re: internal rowId

[Rick James]

MyISAM is implemented thus (simplified, but sufficient for this discussion):
* Rows of data are appended to the data file (.MYD).
* Index(es) are in a separate file (.MYI) and structured in BTree(s).
* The leaf node of an index points to the data via a record number (for Fixed rows) or byte offset (Dynamic).
* If you have no indexes, the data can still be read, essentially by scanning thru the MYD file.
* UNIQUE and PRIMARY are a constraint -- it will check for duplicate keys. There is no diff in the MYD or MYI than for a plain INDEX.
* The PRIMARY KEY, if it exists, is just like any other UNIQUE index.
* The optimizer understands all of the above, looks at all the indexes, and picks zero or one index to perform each SELECT. If it picks zero, the you get a "table scan", as if there were no indexes.

There is no "rowid" as defined by competing database engines.