Miroirs/jorge
1
0
Fork 0
mirror of https://github.com/facundoolano/jorge.git synced 2024-12-25 21:58:28 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions

finish serve post draft

Browse source
This commit is contained in:
facundoolano 2024-03-05 11:12:15 -03:00
parent 28b2d32406
commit 98eb2cd556
1 changed files with 68 additions and 52 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

120
docs/src/blog/a-site-server-with-live-reload.org
View file

@ -9,20 +9,27 @@ draft: true
#+OPTIONS: toc:nil num:1 #+OPTIONS: toc:nil num:1
#+LANGUAGE: en #+LANGUAGE: en
The core of my static site generator is the ~build~ command: take some input files, process them ---render templates, convert other markup formats into HTML--- and write the output for serving to the web. This is where I started with ~jorge~, not only because it was core functionality but because I needed to see the org-mode output as early as possible, to learn if I could expect this project to ultimately replace my Jekyll setup. The core of my static site generator is the ~build~ command: take some input files, process them ---render templates, convert other markup formats into HTML, minify--- and write the output for serving to the web. This is where I started for ~jorge~, not only because it was core functionality but because I needed to see the org-mode output as early as possible, to learn if I could expect this project to ultimately replace my Jekyll setup.
You could say that I had a working static site generator as soon as the ~build~ command was done, but for it to be minimally useful I needed some facility to preview a site while working on it: the ~serve~ command. It could be as simple as running a local file server of the ~build~ output, but ideally it would also watch for changes in the source files and live-reload the browser tabs looking at them. I technically had a working static site generator as soon as the ~build~ command was done, but for it to be minimally useful I needed to be able to preview a site while working on it: a ~serve~ command. It could be as simple as running a local file server of the ~build~ target directory, but ideally it would also watch for changes in the source files and live-reload the browser tabs looking at them.
I was aiming for more than the basics here because ~serve~ was the only non-trivial command of the project: the one with the most Go learning potential ---and the most fun. For similar reasons, I wanted to tackle it as early as possible: since it wasn't immediately obvious how I would implement it, it was here where unknown-unknowns and blockers were most likely to come up. I was aiming for more than just the basics here because ~serve~ was the only non-trivial command of the project: the one with the most Go learning potential ---and the most fun. For similar reasons, I wanted to tackle it early on: since it wasn't immediately obvious how I would implement it, it was here where unknown-unknowns and blockers were most likely to come up.
Once ~build~ and ~serve~ were out of the way, I'd be almost done with the project, the rest being nice-to-have features and UX improvements. Once ~build~ and ~serve~ were out of the way, I'd be almost done with the project, only nice-to-have features and UX improvements remaining.
The beauty of the ~serve~ command was that I could start with a naive implementation and iterate towards the ideal one, keeping a usable command at every step. Below is a summary of that process. The beauty of the ~serve~ command was that I could start with a naive implementation and iterate towards the ideal one, keeping a usable command every step of the way. Below is a summary of that process.
*** A basic file server *** A basic file server
At its simplest, the ~serve~ command consisted of building the site once and serving the target directory with a local server. The standard ~net/http~ package provides [[https://pkg.go.dev/net/http#FileServer][facilities]] for local file servers: The simplest ~serve~ implementation consisted of building the site once and serving the target directory on a local file server. The standard [[https://pkg.go.dev/net/http#FileServer][~net/http~]] package had what I needed:
#+begin_src go #+begin_src go
import (
"net/http"
"github.com/facundoolano/jorge/config"
"github.com/facundoolano/jorge/site"
)
func Serve(config config.Config) error { func Serve(config config.Config) error {
// load and build the project // load and build the project
if err := site.Build(config); err != nil { if err := site.Build(config); err != nil {
@ -38,7 +45,7 @@ func Serve(config config.Config) error {
} }
#+end_src #+end_src
This only required a minor change (based in [[https://stackoverflow.com/a/57281956/993769][this]] StackOverflow answer) to allow omitting the ~.html~ suffix from URLs: I only had to make a minor change (based on [[https://stackoverflow.com/a/57281956/993769][this]] StackOverflow answer) for the server to allow omitting the ~.html~ suffix from URLs so, for instance, ~target/blog/hello.html~ was served at ~/blog/hello~:
#+begin_src go #+begin_src go
type HTMLFileSystem struct { type HTMLFileSystem struct {
@ -58,11 +65,11 @@ func (htmlFS HTMLFileSystem) Open(name string) (http.File, error) {
} }
#+end_src #+end_src
The ~HTMLFileSystem~ above wraps the standard ~http.Dir~ to look for a ~.html~ file when the filename requested isn't found so, for instance, ~target/blog/hello.html~ will be served when receiving a request for ~/blog/hello~. The server setup thus changed to: The server setup thus changed to:
#+begin_src diff #+begin_src diff
- fs := http.FileServer(HTMLFileSystem{http.Dir(config.TargetDir)}) - fs := http.FileServer(http.Dir(config.TargetDir))
+ fs := http.FileServer(http.Dir(config.TargetDir)) + fs := http.FileServer(HTMLFileSystem{http.Dir(config.TargetDir)})
http.Handle("/", fs) http.Handle("/", fs)
fmt.Println("server listening at http://localhost:4001/") fmt.Println("server listening at http://localhost:4001/")
@ -70,7 +77,7 @@ The ~HTMLFileSystem~ above wraps the standard ~http.Dir~ to look for a ~.html~ f
#+end_src #+end_src
*** Watching for changes *** Watching for changes
As a next step, instead of building the site once before running the server I wanted the command to watch the project source directory and trigger new builds every time a file changed. I found the [[https://github.com/fsnotify/fsnotify][fsnotify]] library for this exact purpose; the fact that both Hugo and gojekyll listed it in their dependencies suggested that it was a reasonable choice for the job. As a next step, I needed the command to watch the project source directory and trigger new builds whenever a file changed. I found the [[https://github.com/fsnotify/fsnotify][fsnotify]] library for this exact purpose; the fact that both Hugo and gojekyll listed as a dependency suggested that it was the reasonable choice for the job.
Following [[https://github.com/fsnotify/fsnotify/blob/c94b93b0602779989a9af8c023505e99055c8fe5/README.md#usage][an example]] from the fsnotify documentation, I created a watcher and a goroutine that triggered a ~site.Build~ call every time a file change event was received: Following [[https://github.com/fsnotify/fsnotify/blob/c94b93b0602779989a9af8c023505e99055c8fe5/README.md#usage][an example]] from the fsnotify documentation, I created a watcher and a goroutine that triggered a ~site.Build~ call every time a file change event was received:
@ -92,13 +99,12 @@ func runWatcher(config *config.Config) {
} }
#+end_src #+end_src
Then made the watcher look at changes in the project ~src~ directory: Then made this watcher look for changes in the project ~src/~ directory:
#+begin_src go #+begin_src go
func watchProjectFiles(watcher *fsnotify.Watcher, config *config.Config) { func watchProjectFiles(watcher *fsnotify.Watcher, config *config.Config) {
// fsnotify watches all files within a dir, but non-recursively // fsnotify watches all files within a dir, but non-recursively.
// this walks through the source dir // This walks through the src dir adding watches for each subdir
// adding watches for each found subdir
filepath.WalkDir(config.SrcDir, func(path string, entry fs.DirEntry, err error) error { filepath.WalkDir(config.SrcDir, func(path string, entry fs.DirEntry, err error) error {
if entry.IsDir() { if entry.IsDir() {
watcher.Add(path) watcher.Add(path)
@ -109,15 +115,18 @@ func watchProjectFiles(watcher *fsnotify.Watcher, config *config.Config) {
#+end_src #+end_src
*** Build optimizations *** Build optimizations
At this point I had a useful file server, always responding with the most recent version of the site. But the responsiveness of the ~serve~ command was less than ideal: the entire website had to be processed and copied to the target for any small edit I made on a source file. At this point I had a useful file server, always responding with the most recent version of the site. But the responsiveness of the ~serve~ command wasn't ideal: it processed the entire website for every small edit I made on a source file. I wanted to attempt some performance improvements here, but without introducing much complexity: rather than supporting incremental or conditional builds ---which would have required tracking state and dependencies between files---, I wanted to keep building the entire site on every change, only faster.
I wanted to attempt some performance improvements to the build process, but without introducing much complexity: instead of adding the structure to support incremental or conditional builds, I wanted to try first to keep building the entire site on every change, only faster.
The first cheap optimization was obvious from looking at the command output: most of the work was copying static assets (e.g. images, static CSS files, etc.). So I changed the ~site.Build~ implementation to optionally create links instead of copying files. The first cheap optimization was obvious from looking at the command output: most of the work was copying static assets (e.g. images, static CSS files, etc.). So I changed the ~site.Build~ implementation to optionally create links instead of copying files.
The next thing I wanted to try was to process source files work concurrently. The logic for creating target directories and rendering files was handled by an internal method: The next thing I wanted to try was to process source files concurrently. The logic for creating target directories and rendering files was handled by an internal ~site~ method:
#+begin_src go #+begin_src go
type site struct {
config config.Config
// ...
}
func (site *site) build() error { func (site *site) build() error {
// clear previous target contents // clear previous target contents
os.RemoveAll(site.Config.TargetDir) os.RemoveAll(site.Config.TargetDir)
@ -132,19 +141,18 @@ func (site *site) build() error {
return os.MkdirAll(targetPath, FILE_RW_MODE) return os.MkdirAll(targetPath, FILE_RW_MODE)
} }
// if it's a file render or copy it at the target // if it's a file render or copy it to the target
return site.buildFile(path, targetPath) return site.buildFile(path, targetPath)
}) })
} }
#+end_src #+end_src
This ~site.build~ method walks the source file tree, recreating directories in the target. For non-directory files, it calls another method, ~site.buildFile~, to do the actual processing (rendering templates, converting markdown and org-mode syntax to HTML, "smartifying" quotes, and writing the results to the target files). I wanted the calls to ~site.buildFile~ offloaded to a pool of workers; I found the facilities I needed in a couple of [[https://gobyexample.com/][Go by Example]] entries: This ~site.build~ method walks the source file tree, recreating it at the target. For non-directory files, it calls another method, ~site.buildFile~, to do the actual processing (rendering templates, converting markdown and org-mode syntax to HTML, and writing the results to the target files). I wanted ~site.buildFile~ to run in a worker pool; I found the facilities I needed in a couple of [[https://gobyexample.com/][Go by Example]] entries:
#+begin_src go #+begin_src go
// Runs a pool of workers to build files. Returns a channel // Runs a pool of workers to build files.
// to send the paths of files to be built and a WaitGroup // Returns a channel to send the paths of files to be built
// to wait them to finish processing. // and a WaitGroup to wait for them to finish processing.
Create a channel to send paths to build and a worker pool to handle them concurrently
func spawnBuildWorkers(site *site) (*sync.WaitGroup, chan string) { func spawnBuildWorkers(site *site) (*sync.WaitGroup, chan string) {
var wg sync.WaitGroup var wg sync.WaitGroup
files := make(chan string, 20) files := make(chan string, 20)
@ -162,9 +170,9 @@ func spawnBuildWorkers(site *site) (*sync.WaitGroup, chan string) {
} }
#+end_src #+end_src
The function above creates a buffered channel to receive source file paths, and a worker pool with the size of the amount of CPU cores. Each worker registers itself on a ~WaitGroup~ that can be used by callers to block until all workers finish their work. The function above creates a buffered channel to send source file paths, and a worker pool that reads from it. Each worker registers itself on a ~WaitGroup~ that can be used by callers to block until all workers finish their work.
Then I just needed to adapt the ~build~ function to spawn the workers and send them file paths through the channel, instead of processing them sequentially: Then I just needed to adapt the ~build~ function to spawn the workers and send them file paths through the channel, instead of processing them inline:
#+begin_src diff #+begin_src diff
func (site *site) build() error { func (site *site) build() error {
@ -195,20 +203,15 @@ func (site *site) build() error {
} }
#+end_src #+end_src
the ~close(files)~ call informs the workers that no more work will be sent, and ~wg.Wait()~ blocks execution until all pending work is finished. the ~close(files)~ call informs the workers that no more work will be sent, and ~wg.Wait()~ blocks execution until all pending work is done.
I was very satisfied to see a sequential piece of code turned into a concurrent one with minimal structural changes, without affecting callers of the function I updated. In other languages, a similar process would have required me to add ~async~ and ~await~ statements all over the place. I was very satisfied to see a sequential piece of code turned into a concurrent one with minimal structural changes, without affecting callers of the function that contained it. In other languages, a similar operation would have required me to add ~async~ and ~await~ statements all over the place.
*** Live reload *** Live reload
- intro sse (vs ws) Without having looked into their code, I presumed that the live-reloading tools I had used in the past (~jekyll serve~, [[https://github.com/shime/livedown/][livedown]]) worked by running WebSocket servers and injecting some JavaScript in the HTML files they served. I wanted to see if I could get away with implementing live reloading for ~jorge serve~ with [[https://en.wikipedia.org/wiki/Server-sent_events][Server-sent events]] instead, a slightly simpler alternative to WebSockets that didn't require a dedicated server.
- sse boilerplate
#+begin_src diff Some googling revealed the boilerplate I needed to send events from my Go http server:
fs := http.FileServer(HTMLFileSystem{http.Dir(config.TargetDir)})
http.Handle("/", fs)
+ http.Handle("/_events/", ServerEventsHandler)
#+end_src
#+begin_src go #+begin_src go
func ServerEventsHandler (res http.ResponseWriter, req *http.Request) { func ServerEventsHandler (res http.ResponseWriter, req *http.Request) {
@ -221,8 +224,7 @@ func ServerEventsHandler (res http.ResponseWriter, req *http.Request) {
select { select {
case <-time.After(5 * time.Second): case <-time.After(5 * time.Second):
// send an event to the connected client. // send an event to the connected client.
// data\n\n just means send an empty, unnamed event fmt.Fprint(res, "data: rebuild\n\n")
fmt.Fprint(res, "data\n\n")
res.(http.Flusher).Flush() res.(http.Flusher).Flush()
case <-req.Context().Done(): case <-req.Context().Done():
// client connection closed // client connection closed
@ -232,7 +234,14 @@ func ServerEventsHandler (res http.ResponseWriter, req *http.Request) {
} }
#+end_src #+end_src
The code above will send an empty event every 5 seconds to clients connected to the ~/_events/~ endpoint. After some trial-and-error, I arrived to the following JavaScript snippet for the client side: #+begin_src diff
fs := http.FileServer(HTMLFileSystem{http.Dir(config.TargetDir)})
http.Handle("/", fs)
+ http.Handle("/_events/", ServerEventsHandler)
#+end_src
In this test setup, clients connected to the ~/_events/~ endpoint would receive an event with the ~"rebuild"~ message every 5 seconds. After some trial-and-error, I arrived to the corresponding JavaScript:
#+begin_src html #+begin_src html
<script type="text/javascript"> <script type="text/javascript">
@ -265,9 +274,9 @@ newSSE();
</script> </script>
#+end_src #+end_src
Clients will establish an [[https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events][EventSource]] connection through the ~/_events/~ endpoint, and reload the window whenever a server-sent event arrives. I updated the ~site.buildFile~ logic to inject this script in the header of every HTML file written to the target directory. Clients would establish an [[https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events][EventSource]] connection through the ~/_events/~ endpoint, and reload the window whenever a server-sent event arrived. I updated ~site.buildFile~ to inject this ~script~ tag in the header of every HTML file written to the target directory.
So far I had a working events handler and clients connecting to it. I just needed to update the handler to only send events after site rebuilds triggered by the fsnotify watcher. I couldn't just use a channel to connect both components since every rebuild event needed to be broadcast to all connected clients (there could be more than one open tab at any given moment). I introduced an ~EventBroker~ [fn:1]struct for that purpose, with this API (see the full implementation [[https://github.com/facundoolano/jorge/blob/567db560f511b11492b85cf4f72b51599e8e3a3d/commands/serve.go#L175-L238][here]]): With the code above I had everything in place to send and receive events, and reload the browser accordingly. I just needed to update the http handler to only send events in response to site rebuilds triggered by source file changes. I couldn't just use a channel to connect the handler with the fsnotify watcher, since there could be multiple clients connected at a time (multiple tabs browsing the site) and each needed to receive the reload event; a single-channel message would be consumed by a single client. I needed some method to broadcast rebuild events; I introduced an ~EventBroker~[fn:1] struct for that purpose, with this interface:
#+begin_src go #+begin_src go
// The event broker mediates between the file watcher // The event broker mediates between the file watcher
@ -290,7 +299,9 @@ func (broker *EventBroker) unsubscribe(id uint64)
func (broker *EventBroker) publish(event string) func (broker *EventBroker) publish(event string)
#+end_src #+end_src
The events handler now needed to create a subscription on every client connection, to forward rebuild events through it: See [[https://github.com/facundoolano/jorge/blob/567db560f511b11492b85cf4f72b51599e8e3a3d/commands/serve.go#L175-L238][here]] for the full ~EventBroker~ implementation.
The http handler now needed to subscribe every connected client to receive rebuild events through the broker:
#+begin_src diff #+begin_src diff
-func ServerEventsHandler (res http.ResponseWriter, req *http.Request) { -func ServerEventsHandler (res http.ResponseWriter, req *http.Request) {
@ -307,8 +318,7 @@ The events handler now needed to create a subscription on every client connectio
- case <-time.After(5 * time.Second): - case <-time.After(5 * time.Second):
+ case <-events: + case <-events:
// send an event to the connected client. // send an event to the connected client.
// data\n\n just means send an empty, unnamed event fmt.Fprint(res, "data: rebuild\n\n")
fmt.Fprint(res, "data\n\n")
res.(http.Flusher).Flush() res.(http.Flusher).Flush()
case <-req.Context().Done(): case <-req.Context().Done():
// client connection closed // client connection closed
@ -331,13 +341,13 @@ The watcher, in turn, had to publish an event after every rebuild:
go func() { go func() {
for event := range watcher.Events { for event := range watcher.Events {
fmt.Printf("file %s changed\n", event.Name) fmt.Printf("file %s changed\n", event.Name)
// new src directories could be triggering this event // new src directories could be triggering this event
// so project files need to be re-added every time // so project files need to be re-added every time
watchProjectFiles(watcher, config) watchProjectFiles(watcher, config)
site.Build(*config) site.Build(*config)
+ broker.publish("rebuild") + broker.publish("rebuild")
} }
}() }()
+ return broker + return broker
@ -347,7 +357,8 @@ The watcher, in turn, had to publish an event after every rebuild:
*** Handling event bursts *** Handling event bursts
The code above worked, but not always. Some times, a file change would trigger a browser refresh to a 404 page, as if the new target file wasn't yet written. This was a consequence of single file changes producing many write events, and <it's mentioned in the fsnotify documentation. The solution (also suggested in the doc [LINK]) is to de-duplicate events by adding a delay between event arrival and response. <time.AfterFunc [LINK] helps here The code above worked, but not consistently. A file change would occasionally cause a browser-refresh to a 404 page, as if the new version of the file wasn't written to the target directory yet.
This happened because a single file edit could result in multiple writes, and those in a burst of fsnotify events (as mentioned in the [[https://github.com/fsnotify/fsnotify/blob/v1.7.0/backend_inotify.go#L108-L115][documentation]]). The solution (also suggested by [[https://github.com/fsnotify/fsnotify/blob/c94b93b0602779989a9af8c023505e99055c8fe5/cmd/fsnotify/dedup.go][an example]] in the fsnotify repository) was to de-duplicate events by introducing a delay between event arrival and response. [[https://pkg.go.dev/time#AfterFunc][~time.AfterFunc~]] helped here:
#+begin_src diff #+begin_src diff
@ -379,6 +390,11 @@ func runWatcher(config *config.Config) *EventBroker {
} }
#+end_src #+end_src
The initial build is triggered immediately on setup (~time.AfterFunc(0, ...)~) but subsequent rebuilds are delayed 100 milliseconds (~rebuildAfter.Reset(100 * time.Millisecond)~), canceling previous pending ones.
-----
That's approximately the current implementation of the ~jorge serve~ command, which I used to write this post. You can see the full code [[https://github.com/facundoolano/jorge/blob/28b2d32406c7f4e4f6c3084d521f0123435637c8/commands/serve.go][here]].
** Notes ** Notes
[fn:1] I'm not sure if "broker" is semantically correct in this context, since there's a single event type and is sent to all subscribers. "Broadcaster" is probably more correct, but sounds worse. [fn:1] I'm not sure if "broker" is a proper name in this context, since there's a single event type and it's sent to all subscribers. "Broadcaster" is probably more accurate, but it also sounds worse.