writing-isam-files

Prepare a JSON data file with Fichier, Structure, and Donnees fields, then run write_isam.py with --params-file. The script builds the binary record with Windows-1252 encoding and space padding, then writes it via DhxIsam64.dll using xisam_twritelong.

Pass --op update or --op delete with a Key (logical index name or DLL letter) and KeyFields in the JSON. The script positions via TDF.KEY, reads with tpreadlong mode R, verifies the record matches the key field by field, and only then rewrites or deletes it.

The anti-corruption guardrail locates the .dhsd dictionary declared in the structure, resolves the real table size, and refuses the write with error_code -2 if TailleEnreg differs. This prevents silent corruption, since a .dhfi file can hold multiple tables with different record sizes.

Yes, use the --dry-run flag to validate the structure and data JSON without calling the DLL. You can also run validate_structure.py standalone to check field offsets, sizes, and TailleEnreg constraints before any write.

The most common cause is writing to the wrong physical file path when multiple ERP versions coexist. Check the xwin7 -action synchroauto log to find the .dhfi path actually loaded by the runtime, and verify numeric key fields are right-justified with space padding.

Sensitive files like g3f.dhfi and a5f.dhfi are cached by active ERP sessions, so writes can cause lock conflicts or invisible changes. Close the ERP session before writing, and obtain explicit authorization before update or delete operations on shared standard ERP files.