Manually tell BrowserSync a file has changed through CLI

[hgl]

I'm using watchman to watch files, and once files changed, it will run my build step.

Currently, I ask BrowserSync to watch the build directory, which seems to be inefficient (watch source file change -> run build step -> watch build file change -> refresh browser).

I wonder if it's possible to manually tell BrowserSync to refresh the browser after my build step is done (watch source file change -> run build step -> refresh browser)?

Maybe something like:

$ browser-sync start --server app &
$ browser-sync refresh # browsers will be refreshed after this command is done

or by sending signal to the running browser-sync instance

$ browser-sync start --server app &
$ kill -HUP $!

Or I'm doing it wrong? What's the correct way to use BrowserSync when there is a build step involved?

[shakyShane]

@hgl - we want to implement this asap - but we are stuck on how to send signals between two node processes. BrowserSync/recipes#2 (comment)

If you could shed some light, we'd be extremely greatful.

[hgl]

I guess we can send signals via direct http requests? For example, browser-sync start starts a server, browser-sync refresh talks to the server via http to ask it to refresh. karma uses this architecture for its karma start and karma run commands.

However, I now do realize a downside of manual refreshing: BrowserSync is no longer able to know which files are changed. So it's no longer able to do a quick css update. Every manual fresh must fresh the whole page.

I now find out that by setting reloadDelay to 0. It's actually pretty fast for BrowserSync to notice build file changes. The only problem I encounter is that if the html page is updated by the build step, BrowserSync sometimes refresh the browser to a blank page, manually refresh the browser again fixes the problem. I'm still yet to identify the cause. But fixing this should make BrowserSync very pleasant to use.

[shakyShane]

Ahhh the http idea sounds perfect.

We could also send a bunch of files in the request no problem, why not

# kick it off
browser-sync start --server --port 3000

# for file injection
browser-sync reload --port 3000 --files="core.css,styles.css"

# for normal reloading
browser-sync reload --port 3000

This sounds totally do-able to me - the only downside being that the port must be given each time to identify the running instance

[hgl]

the only downside being that the port must be given each time to identify the running instance

I guess browser-sync reload can read the same config file as browser-sync start does to find the port. If users didn't use a config file then they have to specify the port again.

browser-sync reload --port 3000 --files="core.css,styles.css"

How does this file injection work? To me it looks like telling BrowserSync to only reload those files.

[shakyShane]

Docs coming soon, but this is in 2.6.0 check the release notes