File formats

Text model files

.dat and .cdb are accepted only as text. UTF-8, Windows-1252, and Latin-1 inputs are decoded incrementally. Fixed-width NBLOCK, EBLOCK, and numeric BFBLOCK payloads use the native bulk path. Binary CDB and .db files fail closed.

The deterministic writer emits documented fixed-width NBLOCK and EBLOCK records for bulk mesh state and readable command records for the remaining model. This keeps its output on the same native bulk-read path as Mechanical archives. A write/read round trip preserves the neutral schema rather than original whitespace or record ordering. Numeric parameter arrays use the documented fixed-width *PREAD form; STRING/CHAR arrays use indexed *SET assignments so values wider than a numeric field remain intact. STRING treats its first dimension as published character capacity, while CHAR retains ordinary array-element indexing. Material tables are emitted in six-value MPTEMP/MPDATA slot-continuation groups.

If the imported deck contained a SOLVE boundary, deterministic MAPDL output retains one SOLVE after the preserved solution controls. Model-only archives that did not request a solve do not acquire one.

Results and matrices

The retained MIT reader and native C API cover the publicly documented record layouts used by .rst, .rth, .full, and .emat. Record offsets and lengths are bounds checked before typed decoding. Unsupported compression or an unknown record envelope is reported rather than guessed.

These solver-produced binary families are deliberately read-only. OmniSolver MAPDL Interop does not synthesize result or matrix files; bidirectional conversion applies to the documented text model interface and neutral JSON.

No binary fixture is required by downstream users. Synthetic text examples are the only examples distributed by this project; legally reusable MIT fixtures are restricted to tests and retain their notices.