Improvement of an automatic arrangement for large databases

[aisbergde]

I would like to use dbml also for visualization of references. In principle this works. However, the display is very "vertical". This causes many edges to run in one thick strand. Maybe it is possible to find a way to move the automatic arrangement more to the width, so that the Edges are displayed better. Attached is a dbml file that can be used for testing purposes.

dhw_ColumnReference.zip

[fleipold]

Under the hood the dbml-renderer uses a javascript version of the dot utility of the graphviz suite. It makes use of the .dot-file-format. One could experiment with the alternative algorithms in the graphviz suite to see whether the result is better for a given db.

A .dot file for further experimentation can be easily obtained using the following incantation:

$ dbml-renderer -i example.dbml -o output.dot

[aisbergde]

I know that a dot file is used, but I have no influence on the generated dot file. Should I play with the dot file and then suggest or post some results here?

[aisbergde]

It looks like these are html tables and their placement is "hard coded". This could be OK, if only one table with directly referenced and referencing objects is on the diagram.

Maybe there is a possibility to change the color of a relation when hoovering the mouse? I think SVG supports this, and maybe also dot will support this to generate the svg from these dot files?
Then you could interactively recognize the relations to the columns.

Here it is hard to identify column linage:

[d3x0r]

Under the hood the dbml-renderer uses a javascript version of the dot utility of the graphviz suite. It makes use of the .dot-file-format. One could experiment with the alternative algorithms in the graphviz suite to see whether the result is better for a given db.

A .dot file for further experimentation can be easily obtained using the following incantation:

$ dbml-renderer -i example.dbml -o output.dot

-o output.dot writes SVG into a file called output.dot...

I too would like better separation - maybe even a much wider canvas?

(example output, and a .dbml input)
output.zip