[core][dag] Add ascii based CG visualization

[Bye-legumes]

Why are these changes needed?

Add ascii based CG visualization so we can have better view of the CG.

Related issue number

Close
#48190

Example

codes

import ray
from ray.dag.input_node import InputNode
from ray.dag.output_node import MultiOutputNode

ray.init()
@ray.remote
class Actor:
    @ray.method(num_returns=3)
    def return_three(self, x):
        return x, x + 1, x+2

    def echo(self, x):
        return x
    
    @ray.method(num_returns=2)
    def return_two(self,x):
        return x, x + 1

a = Actor.remote()
b = Actor.remote()
with InputNode() as i:
    o1, o2, o3 = a.return_three.bind(i)
    o4 = b.echo.bind(o1)
    o5 = b.echo.bind(o2)
    o6, o7 = b.return_two.bind(o3)
    dag = MultiOutputNode([o4, o5, o6, o7])

compiled_dag = dag.experimental_compile()

# Get the DOT source
res = compiled_dag.visualize(format='ascii')
print(res)






#compiled_dag = dag.experimental_compile()
# Call the visualize method
#compiled_dag.visualize(filename=filename, view=False)
#res = compiled_dag.execute(1)
#print(ray.get(res))
# Assert that the visualization file exists
#assert os.path.exists(f'{filename}.png'), "Visualization file was not created."

# Clean up
#compiled_dag.teardown()

output

Nodes Information:
            0 [label="Task 0  InputNode"]
            1 [label="Task 1  Actor: 54777d... Method: return_three"]
            2 [label="Task 2  ClassMethodOutputNode[0]"]
            3 [label="Task 3  ClassMethodOutputNode[1]"]
            4 [label="Task 4  ClassMethodOutputNode[2]"]
            5 [label="Task 5  Actor: c927c9... Method: echo"]
            6 [label="Task 6  Actor: c927c9... Method: echo"]
            7 [label="Task 7  Actor: c927c9... Method: return_two"]
            8 [label="Task 8  MultiOutputNode"]
            9 [label="Task 9  ClassMethodOutputNode[0]"]
            10 [label="Task 10  ClassMethodOutputNode[1]"]

            Edges Information:
            0 ---> 1 [label=SharedMemoryType]
            1 ---> 2 [label=SharedMemoryType]
            1 ---> 3 [label=SharedMemoryType]
            1 ---> 4 [label=SharedMemoryType]
            2 ---> 5 [label=SharedMemoryType]
            3 ---> 6 [label=SharedMemoryType]
            4 ---> 7 [label=SharedMemoryType]
            5 ---> 8 [label=SharedMemoryType]
            6 ---> 8 [label=SharedMemoryType]
            9 ---> 8 [label=ChannelOutputType]
            10 ---> 8 [label=ChannelOutputType]
            7 ---> 9 [label=SharedMemoryType]
            7 ---> 10 [label=SharedMemoryType]

            Experimental Graph Built:
            0:InputNode
            |
            1:Actor_54777d:return_three
            |---------------------------->|---------------------------->|                                                  # noqa
            2:Output[0]                   3:Output[1]                   4:Output[2]                                        # noqa
            |                             |                             |                                                  # noqa
            5:Actor_c927c9:echo           6:Actor_c927c9:echo           7:Actor_c927c9:return_two                          # noqa
            |                             |                             |---------------------------->|                    # noqa
            |                             |                             9:Output[0]                   10:Output[1]         # noqa
            |<----------------------------|-----------------------------|-----------------------------|                    # noqa
            8:MultiOutputNode

Checks

• [√] I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• [√] I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • [√] Unit tests
  • [√] Release tests
  • This PR is not tested :(

[Bye-legumes]

@ruisearch42 Hi, can you check this. I just add another method based on ascii and the previous visualization keeps the same

[ruisearch42]

Thanks for the contribution.

I think ASCII visualization only scales to a certain graph size and complexity and would like to get an idea what types of graphs work well for this implementation and what would break.

[Bye-legumes]

Hi, @ruisearch42, I just made some modification based on your comments. I think it looks better now. While I prefer to return the sting instead of print so we can run the tests.

[ruisearch42]

I just had some ideas to make the viz easier to read. After that I will probably have a final round of review for merging.

[Bye-legumes]

I just had some ideas to make the viz easier to read. After that I will probably have a final round of review for merging.

I just make some fix, can you check these? Thx!

[ruisearch42]

I will take a look at the PR today.

[ruisearch42]

overall LG

[Bye-legumes]

@ruisearch42 HI, I just fix some issues you mentioned and also explained some doc, Can you check it? Thx!

[Bye-legumes]

@ruisearch42 I just changed the return and unified them, So by default, for non-ascii, it will store the figure in disk and also return the string. In test, we just need to clean the file generated.

[Bye-legumes]

@ruisearch42 I think it's OK now!

[ruisearch42]

Thanks for the change!