Help/IRC - Forum - Manual - What's New - Download
Install: Linux, Windows, Mac
Build: Linux, Windows, Mac, Source, Checkout, Doxygen
Exodus: The Next Revolution in Multivalue?
Many people have dreamed of migrating their classic multivalue database applications and/or skills into mainstream programming environments. Until now however, it has been impossible. Where migration has actually been done, it has always meant that applications and skills have had to be redeveloped. The usual result is that multivalue applications are simply kept in operation and interfaces added to other systems. This isn't migration though, it is co-existence.
Exodus is freely licenced implementation of a unique concept that completely reincarnates multivalue database technology in a mainstream environment with:
- no proprietary/legacy database
- no proprietary/legacy language/compiler
- no proprietary/legacy runtime engine
For those who know the history of the multivalue database world, Exodus promises its next revolution. First, multivalue stopped providing an operating system, next its data store became replaceable with mainstream databases. Now, with Exodus, even the Basic language/compiler has been eliminated.
Using only widely accepted tools should ease adoption of Exodus (and therefore multivalue based applications) by the vast majority of companies and individuals who are not familiar with multivalue technology and are more comfortable investing in mainstream products.
Universal Programming Style
To enable a smooth migration of an application, all the underlying principles of multivalue databases must be retained and the code must be as readable and as writable as is was before. This is what Exodus provides. A change of syntax is unavoidable since Exodus uses a mainstream compiler, but it does not take long to learn and large bodies of existing code can be converted reliably and routinely.
Exodus uses PostgreSQL for all data storage while making it appear just like a completely traditional multivalue database - and without requiring any knowledge of SQL. All data definition is done in what appear to programmer/dba to be classic multivalue dictionary files.
Exodus supports various dictionary formats, the commonest iconv/oconv functions and has a complete LIST/ENGLISH/ACCESS replacement that can output either text or html.
In the very near future all multivalue data will also be *readable* via Postgresql's standard SQL interfaces. This is without any additional work on behalf of the programmer/dba beyond the maintenance of classic multivalue dictionary files.
All Exodus strings are Unicode and ordering is done using operating system collators. Internally, Exodus is UCS-2 on Windows 32bit and UTF-32 on Unix/Linux/Mac/Win64. All file i/o is locale/codepage based or UTF-8.
Exodus has a binary installer for Windows and OSX. Unix/Linux/Mac can be installed using configure/make/make install.
Exodus has no dependencies other than the obvious. If you want to write programs in Exodus, you need a C++ compiler. If you want to access or update multivalue data, you need Postgresql 8.1+, 9.0+.
Exodus's command line utilities are all written in Exodus/C++ providing both portability and samples of real programming using Exodus.
If you want to build Exodus from source, you need Boost libraries. Portability to other platforms is forecast to be high. The code base is quite small and comprehensible being mostly implementation of a single class called "var".
Cross Platform Development
To ease the life of Exodus programmers on all platforms, Exodus provides edic and other commands to edit/compile/catalog programs from any standard operating system command console. edic wraps the ubiquitous nano editor with color coding of exodus/c++ syntax and "jump to first error" after compilation errors. You can configure your own editor where desired. The Windows installer includes nano.
For traditional multivalue programmers, Exodus remains in the familiar world of "compiling" source code and executing "object code". Modern C++ compilers and debuggers are very sophisticated these days. When compiled in the default debug mode, Exodus provides a source level backtrace for run-time errors like "variable not assigned", "non-numeric data" etc. All exodus programs can be interrupted/resumed.
The massive range of commercial and open source tools is available to assist development in C++.
Deployment of Applications
In the standards-based world that Exodus inhabits, there is a huge variety of tools to deploy applications.
Exodus has a tiny footprint, is thread-safe and can produce standalone programs or be linked into utilities, libraries or embedded projects - with or without database functionality. The possibilities are very broad. The Exodus library can be maintained by any average C++ programmer, there is nothing particularly difficult about its implementation. The best things are always simple, like the original Pick database - which looks like living on in yet another incarnation in Exodus.
Genesis - Looking Back
Exodus has been developed to date by NEOSYS to serve as the new back-end platform for the next generation of their vertical market package. It is nominally 90% complete and the input of multivalue professionals is sought before the API is frozen and put into production.
Roadmap - Looking Forward
- Completion of Exodus Programmer's Manual
- mv-programmer-friendly way of doing labelled common
- Sorting and indexing on multivalues
- oconv/iconv of dates, time and numbers in any locale
Possible future development
- Complete data visibility via SQL (automatic from multivalue dictionary)
- Binary build for main Linux distros
- Move everything onto sourceforge
- Move unit tests into subversion
- mv programmer friendly db management utils file/account-create/save/restore/
- Ease to use dictionary editor instead of plain edit?
- F correlatives
- Port the Postgres plugin to other databases
Announcement and discussion of project []
How does Exodus handle Unicode
Why is Exodus dependent on Boost and not Qt?
Why does Exodus use autotools and not Cmake?
Getting the right to commit to the subversion repository
Unorganised notes about programming in Exodus
Comparison of "Key/Value" databases v. SQL databases
include\exodusfuncs.h lists the functions that simulates pickish syntax as close as can be achieved. The functionality however should be identical.
include\exodusmacros.h lists a few keywords to simplify programming for application programmers not interested in programming in a more OO style.
include\mv.h lists the massive var class definition that defines how to write pick in oo object.method() style syntax.