We probably should be validating that the folder is a directory otherwise the is_directory will be not valid.

for sure will add this check

We do have logic that walks a full directory. https://github.com/streamlit/streamlit/blob/develop/lib/streamlit/watcher/path_watcher.py#L143

It seems to just create a watch on the path, and that can allow for . https://github.com/streamlit/streamlit/blob/develop/lib/streamlit/watcher/path_watcher.py#L119

Perhaps we can just do a watch on the path itself. Perhaps we can modify _register_watcher. We can test if the glob pattern needs to be set. This would reduce watchers generated and perhaps simplify the logic.

for sure i'll look at incorporating the watch_dir function and try to use the glob pattern to make watching paths easier

Created a seperate function for this as I didn't want to interfere with existing logic which uses the on_file_changed callback, and make it more explicit that this callback can be called for changes to directories

I didn't need to create this function.

removed this function

as requested using the watch_dir with the usage of glob_pattern to search all subdirectories and types of files to walk the directory files

I think this will simplify if you update the entire try block to just be this.

kwargs = {}
if is_directory:
    kwargs["glob_pattern"] = "**/*"

wm = WatchedModule(
    watcher=PathWatcher(filepath, self.on_file_changed, **kwargs),
    module_name=module_name,
)
self._watched_modules[filepath] = wm

Apologies if I was confusing, but my comment was on watch_dir was to borrow the logic from it. Looking at the function, it's simply creating a watcher with the glob pattern, and we already have the watcher, so we just need to specify the glob pattern if it's a directory. self.on_file_changed will have the correct file path because the watcher will supply the directory path, and not the actual file that changed. That's fine with us because we just want to rerun the page if we detect any change.

As an additional cleanup, I might consider modifying the name of on_file_changed to on_path_changed.

applied this

This needed to be done since the event_based_path_watcher only checks for abs_changed_paths, this approach checks if the path that has changed is a common ancestor of the parent path. making it simple to check for for changes to a directory no matter the file under the directory

So I was able to get everything working with removing this code. Seems to work fine in my tests.

@kmcgrady I can remove the check for if changed_path_info is None, but if I remove the for loop going over the watched_paths.items()and checking if they share common path then it fails to update the app there is a change in the directories being watched.

You shouldn't need to specify the type_. I think it's a little weird, but I think it should be str because the values are strings, but we will accept multiple strings for the paths. The fact that the default val is a list hints at taking multiple values, and str is inferred by default.

So I was able to get everything working with removing this code. Seems to work fine in my tests.

I didn't need to create this function.

I think this will simplify if you update the entire try block to just be this.

kwargs = {}
if is_directory:
    kwargs["glob_pattern"] = "**/*"

wm = WatchedModule(
    watcher=PathWatcher(filepath, self.on_file_changed, **kwargs),
    module_name=module_name,
)
self._watched_modules[filepath] = wm

Apologies if I was confusing, but my comment was on watch_dir was to borrow the logic from it. Looking at the function, it's simply creating a watcher with the glob pattern, and we already have the watcher, so we just need to specify the glob pattern if it's a directory. self.on_file_changed will have the correct file path because the watcher will supply the directory path, and not the actual file that changed. That's fine with us because we just want to rerun the page if we detect any change.

As an additional cleanup, I might consider modifying the name of on_file_changed to on_path_changed.