fix: make flow name retrieval efficient

[lemorage]

This PR will make cocoindex ls command way more efficient than before, and fix #412.

Context

In the previous versions, we tried to access each Flow’s name property in the list comprehension:

current_flow_names = [fl.name for fl in flow.flows()]

The name property of a Flow is defined as:

@property
def name(self) -> str:
    return self._lazy_engine_flow().name()

Each call to self._lazy_engine_flow() triggers the creation of the underlying _engine.Flow if it hasn’t been built yet. Till now, you can see we have done a lot of unnecessary jobs here, since only we need here is to retrieve flow names.

Solution

We use the helper now to quickly get flow names as a list.

def flow_names() -> list[str]:
    """
    Get the names of all flows.
    """
    with _flows_lock:
        return list(_flows.keys())

Result

Compared to the result in #412, now we've made it 10 times faster.

 cocoindex  time uv run main.py cocoindex ls
CodeEmbedding [+]
DocsToKG [+]
TextEmbedding [+]
GoogleDriveTextEmbedding [+]

Notes:
  [+]: Flows present in the current process, but missing setup.
uv run main.py cocoindex ls  2.01s user 0.47s system 68% cpu 3.647 total

[badmonster0]

Thanks for submitting the fix!