Greetings!
I would like to share this HDF5 utility project to generates an HDF5 container with random layout and content. In addition to its main function, with graphviz it can visualise the generated prufer sequence based tree.
The MIT licensed project is meant to be used in a hackish way, customisation, pull requests, etc are highly encouraged.
best wishes: steven
âPrĂźferâ also happens to mean âtesterâ in German. And what a tester this is! A real pressure cooker for the HDF5 library 
If using h5rnd to âstress testâ a UI, for example, long data object names, high rank, large dimensions, deeply nested datasets, etc. would be very useful. (In fact, itâs anything whose representation impacts the geometry of canvas allocations.)
Would it be possible to include an illustrative HDF5 file in the repo for a quick look at the generated content?
Maybe this an area that would require some customization. But it seems more promising than having relatively static generators, or scouring the internet for âgoodâ HDF5 files.
Thanks for the interest! I updated the h5rnd repository with the output for:
h5ls -r tree.h5
dot see manual for detailsdot renderer: dot -Tsvg tree.gv -o tree.svg
enjoy!
HDF5 container h5ls -r tree.h5
/ Group
/AkWTsmZmlndfbRhKWywY Dataset {1000}
/luSQtFEs Group
/luSQtFEs/EcdhmBaFpNoplYIqUGHuVZj Group
/luSQtFEs/EcdhmBaFpNoplYIqUGHuVZj/ORmqQLNnWtDSQhfexAyrczVolGPY Dataset {1000}
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv Group
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/BtuejM Dataset {1000}
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/YSdbCOyP Group
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/YSdbCOyP/DNrnKgcxo Group
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/YSdbCOyP/DNrnKgcxo/CudEjqyaFtBGqcrFenMUhA Dataset {1000}
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/YSdbCOyP/VDorTtsiEIwtZ Dataset {1000}
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/YSdbCOyP/fuLsrYEurcvdOXWickWKFXLyMGY Dataset {1000}
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/qyHzilzoCyZlTaMcqZQ Dataset {1000}
/luSQtFEs/pOnCmrTKREyihDZBHmDyJDpiLv/wfXkhrapzAB Dataset {1000}
/luSQtFEs/yGnBRCmSxyrPv Group
/luSQtFEs/yGnBRCmSxyrPv/bZPYDqZmhSXwunlyCwGMLu Group
/luSQtFEs/yGnBRCmSxyrPv/bZPYDqZmhSXwunlyCwGMLu/PnHPAPgjWqfkhBHeR Dataset {1000}
/luSQtFEs/yGnBRCmSxyrPv/bZPYDqZmhSXwunlyCwGMLu/QmNJYWKrrEXsqVxDAqkFQRPVHwaD Dataset {1000}
/luSQtFEs/yGnBRCmSxyrPv/rtaEOCMRrqKzAcRBEHQjFQ Dataset {1000}
/snWTGjtKQva Dataset {1000}
/zzxgLfDdtDtViDnSLtrEzVpYKVZzr Dataset {1000}
graphviz: cat tree.gv sets defaults for datasets, then lists the folders, use dot directed graph for rendering, see: Makefile for details
digraph prufer {
node [shape=note color=orange fontcolor=purple fontname="times bold" fillcolor=violet width=0.01 height=0.1 fontsize=9.0 labelloc=b style=bold];
1 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
2 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
3 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
4 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
5 [shape=cylinder color=purple fillcolor=orange style=filled label=HDF5 weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
6 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
7 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
8 [shape=folder color=purple weight=2.0 width=0.75 height=0.5 fontsize=14.0 labelloc=c]
6 -> 0;
2 -> 9;
1 -> 2;
4 -> 10;
3 -> 4;
7 -> 11;
1 -> 12;
6 -> 13;
8 -> 14;
8 -> 15;
7 -> 8;
3 -> 7;
5 -> 16;
5 -> 17;
6 -> 18;
1 -> 19;
6 -> 1;
3 -> 6;
5 -> 3;
5 -> 20;
}
This is excellent. That helped a lot!
Would it be a reasonable extension to increase the diversity of the datatypes, and randomize array dimensions and dataset ranks? Also perhaps chunked datasets?
If so, I can consider a fork. (It would be an Windows/VS2017 build, as Iâm set up for only Windows platforms.)