This extension contains a reader and writer for HDF5 files. The included sample processes (HDF5 Repository/processes) show how to join a set of HDF5 tables from one file, how to split a set of tables into two sets and how to write an example set to an HDF5 file.
Not all HDF5 formats are supported; entries that are not datasets will be ignored, as well as attributes of type compound (table in tables), opaque, bit field, reference and no class.
Since version 5.4.3, new attribtues can be added and old values be manipulated. This might increase memory usage but allows usage with all operators.
- Operationg system: Windows, Linux, Solaris, MacOs
- Java 7
- RapidMiner 5.3 or higher
- possibly administrator rights
For questions and problems, please write an Email to Jan Czogalla
Anmerkung: Die Marketplace Version erkennt das Betriebssystem automatisch, unterstützt aber nur die fünf oben genannten Betriebssysteme.
- Fixed an issue when dragging HDF5 reader operator into a process
- Changes for compatibility with 5.3 and 7.2+
- Small changes
- Improved error handling of nonexistent or faulty data sets
- Small changes due to RapidMiner 7
- Fixed finding a directory for required HDF libraries
- Added parameter to HDF5Writer to control handling of nominal attributes
- Attribute roles are preserved after writing
- Fixed small bugs
- Added the HDF5Writer operator
- Fixed an issue with the preload mechanism
- Used sparse arrays to represent data rows
- Reworked the data structure of new attributes
The source code is available via GIT here.
Q: On Windows, why does the HDFReader operator not show up after installation?
A: When the extension is started, it creates a folder and copies the native HDF libraries there. After installation or update of the extension, run RapidMiner once as administrator to allow the copying of the files.
Q: On Windows, why does RapidMiner show the HDF5 extension to be incompatible?
A: It is the same problem as in the question above. Run RapidMiner as administrator once to solve this problem.