Prepare
- Upload HP 3000 / IMAGE files
Guide
IMAGE Database (later TurboIMAGE) is a legacy network-model database system for the HP 3000 computer platform, organizing data into master datasets with hashed key access and detail datasets linked by chain-based relationships.
Planned Support
- Image Database versions E.x through F.x
- Master and detail dataset extraction
- Master-detail relationship mapping
- Record structure recovery
What You Get Out
Once the parser ships, DataMeans will extract your data into multiple modern formats:
| Output | Description |
|---|---|
csv/{TableName}.csv | One CSV file per table with all row data |
xlsx/{TableName}.xlsx | Excel workbook per table |
xls/{TableName}.xls | Legacy Excel format per table |
json/{TableName}.json | JSON array of records per table |
json/{TableName}.jsonl | Newline-delimited JSON (streaming-friendly) |
postgres.sql | PostgreSQL CREATE TABLE + INSERT statements |
schema/schema-graph.json | Relationship graph for visualization |
schema/er-model.json | ER model for diagram tools |
report.json | Structured extraction report |
report.md | Human-readable extraction summary |
File Requirements
- Database root file (schema and relationship definitions)
- Master and detail dataset files
Current Status
Parser development is in the planning phase. MPE/iX, the HP 3000 operating system, is no longer supported, making testing and validation challenging.
Technical Notes
This is an MPE/iX-specific proprietary format. The operating system is extinct, which creates significant barriers for development and testing. Specialized expertise in HP 3000 systems is required.
Last updated: January 2026
Technical reference
Overview
The HP 3000 was a family of business minicomputers from Hewlett-Packard, first shipped in 1972 and running the MPE (Multi-Programming Executive) operating system. IMAGE — later named TurboIMAGE, TurboIMAGE/XL, and IMAGE/SQL — was the database management system bundled with the platform and was the first DBMS included with a business-class minicomputer. IMAGE uses a network data model: data is stored in master and detail datasets, master entries are located by hashing the key item value, and related detail entries are connected through pointer chains rather than relational joins.
History and Background
- 1971: HP announces the HP 3000; the first system ships in November 1972.
- 1972: IMAGE/3000 database management system introduced for the HP 3000.
- 1973: Early systems are withdrawn to fix performance and stability problems; a redesigned HP 3000 is reintroduced in November 1973.
- 1976: HP 3000 Series II ships in May, the first widely successful model of the line.
- 1983: MPE V operating system released in December.
- 1984: HP 3000 Series 37 introduced, the first model in the line able to run in an office without special cooling.
- 1987: First 32-bit PA-RISC HP 3000s introduced, running MPE XL with TurboIMAGE/XL.
- 1992: MPE renamed MPE/iX after POSIX compatibility is added; HP announces IMAGE/SQL in December.
- 2001: HP announces the end of the HP 3000 line in November; final sales end in 2003.
- 2002: MPE/iX 7.5, the final release of the operating system, ships.
- 2010: HP ends HP 3000 support on December 31, after 38 years.
File Format Specifications
IMAGE databases are stored as privileged MPE disk files generated from a schema by the DBSCHEMA processor.
File Extensions:
- No specific extensions - MPE files
- Root file: carries the database name itself (e.g., ORDERS)
- Dataset files: root file name plus a two-character suffix (e.g., ORDERS01, ORDERS02)
- B-tree index files: POSIX-named with an
.idxsuffix (e.g.,ORDERS02.idx) - Jumbo dataset chunks: a chunk-control file plus POSIX-named chunk files suffixed
.001,.002, and so on (e.g.,ORDERS01.001)
File Structure:
- Root file: single-extent file holding the database definition
- Master datasets: keyed entry points with hashed access
- Detail datasets: data entries linked into chains by search item
- Chains: pointer-linked entry lists with chain heads stored in master entries
- Synonym chains: resolve hash collisions within master datasets
- Media records: each record stores one data entry plus its pointer information; record length is constant within a dataset file
- Blocks: media records are grouped into blocks, each corresponding to one MPE file record, with the blocking factor set by the schema processor
- User labels: every data file carries a label holding structural pointers and counters used for dynamic storage allocation
- Dynamic expansion: datasets defined with initial capacity, increment, and maximum capacity values grow automatically during DBPUT
Key Components:
- Manual master datasets: key entries added explicitly by applications
- Automatic master datasets: key entries maintained by IMAGE itself
- Detail datasets: hold data entries reached through chains
- Intrinsics: library calls such as DBOPEN, DBFIND, DBGET, DBPUT, DBUPDATE, DBDELETE, DBLOCK, and DBCLOSE
- Transaction intrinsics: DBXBEGIN, DBXEND, and DBXUNDO delimit dynamic transactions with roll-back support
- Utilities: DBSCHEMA, DBUTIL, and the QUERY/3000 inquiry facility
- Maintenance utilities: DBSTORE, DBRESTOR, DBUNLOAD, DBLOAD, and DBRECOV handle backup, reload, and recovery
System Limits (TurboIMAGE/XL):
- 240 datasets and 1,200 data items per database; 255 data items per dataset
- 64 detail paths per master dataset; 16 paths per detail dataset
- Maximum entry length 4,756 bytes; item length 4,094 bytes; block length 5,120 bytes
- Up to 2,147,483,647 entries per dataset and per chain; 99 chunks per jumbo dataset; 396 GB per dataset
- B-tree index keys up to 252 bytes
Data Types and Structures
| Type | Description | Storage |
|---|---|---|
| I (Integer) | Signed binary integer in two's complement form | Halfwords (16 bits) |
| J (Integer) | Signed integer conforming to COBOL COMPUTATIONAL limits | Halfwords (16 bits) |
| K (Unsigned) | Absolute binary quantity; no negative values | Halfwords (16 bits) |
| R (Real) | HP 3000 floating-point number | Halfwords (16 bits) |
| E (Real) | IEEE floating-point number | Halfwords (16 bits) |
| P (Packed) | Packed decimal number | Nibbles (4 bits) |
| Z (Zoned) | Zoned decimal number | Bytes |
| X (Alphanumeric) | Unrestricted ASCII character string | Bytes |
| U (Uppercase) | ASCII string containing no lowercase characters | Bytes |
Network Model:
- Master datasets provide one keyed entry per unique key item value
- Detail entries sharing a search item value are linked into a chain
- Chains can be kept sorted on a designated sort item
- One path per detail dataset can be designated the primary path, whose chain order DBUNLOAD preserves in chained mode
- Hashing locates master entries directly from key values
- Optional B-tree indexes on master datasets extend DBFIND with sorted and partial-key retrieval
Version Differences
| Version | Year | Key Changes | Compatibility |
|---|---|---|---|
| IMAGE/3000 | 1972 | Initial network-model DBMS bundled with MPE | Classic 16-bit HP 3000 |
| TurboIMAGE | 1985 | Raised dataset capacities and internal limits | MPE V |
| TurboIMAGE/XL | 1987 | Native PA-RISC version (C.00.00) | MPE XL |
| IMAGE/SQL | 1993 | SQL interface over existing TurboIMAGE databases | MPE/iX |
| TurboIMAGE/XL C.07.04 | 1997 | B-tree indexes; root file version level C4 | MPE/iX 5.5 |
| TurboIMAGE/XL C.07.10 | 1998 | Maximum dataset size raised from 40 GB to 80 GB by widening block addressing to 2^24 − 1 blocks | MPE/iX 6.0 |
| TurboIMAGE/XL C.09.00 | 2001 | Entrynumber format (32-bit record numbers) via $CONTROL LARGESET; root file version level C6 | Entryname databases unchanged; datasets over 80 GB require entrynumber format |
Compatibility Notes:
- Database files are specific to MPE and are not directly portable to other filesystems
- Root file version levels mark structural features: C2 (base), C3 (jumbo datasets), C4 (B-tree indexes), C6 (entrynumber format)
- Jumbo datasets, introduced with MPE/iX 5.5 and patched into 5.0, are skipped by DBSTORE; they are backed up with the STORE command using POSIX names
- IMAGE/SQL reads and updates existing TurboIMAGE databases without conversion
- Third-party indexing products (Omnidex, Superdex) supplement IMAGE access paths
- Marxmeier Eloquence offers a TurboIMAGE-compatible API on HP-UX and Windows
- The Stromasys CHARON-HPA/3000 emulator (2012) provides full HP 3000 hardware emulation on x86-64 Linux servers running MPE/iX
Technical References
- HP 3000 (Wikipedia)
- IMAGE (database) (Wikipedia)
- HP Multi-Programming Executive (Wikipedia)
- TurboIMAGE/XL Database Management System Reference Manual (HP 30391-90012)
- HP Computer Museum: IMAGE/3000
To learn how to use this format with DataMeans, see the User Guide.