marc22alain / algorithms Goto Github PK

Multiple nodes of the same name will be displayed, while only the most recently produced one will be valid.
This only occurs if not clicking on the "Clear all elements" in the top window menu between the loading of each graph.

The expected behaviour of the "load graph" button is therefore ambiguous: should it allow for extending the current graph, or should it clear the current one before creating the new one?

The algorithm implementations provide no signals to graph_maker that they are done, but additional calls to their doStep() methods may raise a variety of exceptions.
(Also, not all exceptions are handled in doStep().)

It is therefore unclear to the user that the algorithm has in fact proceeded to completion and not failed with some error.

The CLRS algorithm explores adjacent vertices, so the implementation does too (neighbouring vertices), without recording any data relatable to the edges themselves. Therefore, when drawing the graph, there is no property of the edges that can direct how they should be drawn.
CLRS do show how edges are explored, with step-by-step illustrations in the text-book, but their pseudo-code does not support that either.

...and negative cycle detection.

This can become a new check for the unit tests. Should be able to apply different algorithms on same input and get predictably different results.

The user can choose an algorithm at any time, either after the graph is complete, during its production, or when it is empty. Currently, when the user chooses an algorithm in the top window menu, it is initialized with the graph in that current state. Later user additions to the graph do not re-initialize the graph.
Other case: the user has used the algorithm to process a graph, then clears all and begins a new one. The clearAll method will reinitialize the algorithm with an empty graph.
The only time that we really know that the user is done producing the graph is when they first click on the "Do step..." button.

Consider its context of use, to determine what features the class should offer.

Want to be able to compare it to Kruskal.

Implement Ford-Fulkerson; implement min-cut.

Implement the 'doStep()' method and implement a drawing procedure that well explains the single-source-shortest-paths progression.
Also determine how the user can define the source vertex in graph_maker.