Goal is to embed a directory tree in my binary, then copy it to a user directory later. It consists mostly of text files. I am attempting to display contents of the files in that subdirectory within the project, named .config (hence the use of go:embed all).
As I understand Go embedding, the following should work outside its own directory, but when executed, lists the name of the first
file in the directory tree but cannot open it or display its contents. Within its own project directory it works fine.
//go:embed all:.config
var configFiles embed.FS
func main() {
ls(configFiles)
}
func ls(files embed.FS) error {
fs.WalkDir(files, ".", func(path string, d fs.DirEntry, err error) error {
if err != nil {
return err
}
if !d.IsDir() {
show(path) // Display contents of file
}
return nil
})
return nil
}
(Complete code at https://go.dev/play/p/A0HzD0rbvX- but it doesn't work because there's no .config directory)
The program walks the embedded file system, but opens files using the operating system. Fix by opening the file in the file system.
Pass the file system to show:
if !d.IsDir() {
show(files, path) // <-- pass files here
}
Open the file using the file system:
func show(files fs.FS, filename string) { // <-- add arg here
f, err := files.Open(filename) // <-- FS.Open
Related
myprogram/
|
|-main.go
|-dir1/
|-data/
|-datafile.json
|-runner.go
|-runner_test.go
In runner.go, I have a simple function that reads the datafile.json. Something like
func GetPayload() (string, err) {
dBytes, dErr := ioutil.ReadFile("dir1/data/datafile.json")
if dErr != nil { return nil, dErr}
return dBytes, nil
}
I'm using Go in a Lambda with a structure similar to above. When the Lambda runs in its actual environment, it starts at main.go, and then invokes GetPayload() from runner.go. However, I have a test in a simple worker node machine in runner_test.go that also hits GetPayload() .
During "normal" execution (from main.go) - this works OK. However, when GetPayload() is invoked from runner_test.go, it errors, saying
open dir1/data/datafile.json no such file or directory
This makes sense, because during the test, the working directory is the directory that houses runner_test.go, which is data/, so there is no dir1 as a child of it. I've been trying to play with using os.Getwd() and getting the paths from there like:
pwd, _ := os.Getwd()
dBytes, dErr := ioutil.ReadFile(pwd + "dir1/data/datafile.json")
But again, that won't work, because for runner_test.go pwd is user/myname/myprogram/dir1, but from main.go, it turns up as user/myname/myprogram.
Any idea how I can find and open datafile.json from within GetPayload() in any environment? I could pass an optional parameter to GetPayload() but if possible, it'd be great to avoid that.
If the file is static (meaning that it doesn't need to change after you build the program), you can embed it into the built program. This means you no longer have to worry about run-time file paths.
import (
"embed"
)
//go:embed data/*
var dataFiles embed.FS
func GetPayload() (string, err) {
dBytes, dErr := dataFiles.ReadFile(dataFiles, "data/datafile.json")
if dErr != nil { return nil, dErr}
return dBytes, nil
}
Now the files in your data/ directory are embedded in this variable dataFiles which acts as a read-only file system.
For more info:
Read more about embed in the package documentation overview.
Read my answer about "when to use embed"
For data files that your program needs during runtime, either use a fixed directory and refer to that, or accept a command line argument or some sort of configuration that tells you where the file is.
When running unit tests, the wd is the directory containing the test file. One convention is to use a testdata/ directory under the directory containing the test, and put all data files there. That way you can refer to that file from the test by using testdata/datafile.json.
You can use a copy of the file you need during runtime as your test file, or you can use a symlink from the runtime data file to the test file under the testdata/ dir.
For data files that your program needs during runtime, either use a fixed
directory and refer to that
Someone made this suggestion, which I agree with. To that end, you can use
something like this:
package main
import (
"os"
"path/filepath"
)
func main() {
d, err := os.UserCacheDir()
if err != nil {
panic(err)
}
d = filepath.Join(d, "file.json")
f, err := os.Open(d)
if err != nil {
panic(err)
}
defer f.Close()
os.Stdout.ReadFrom(f)
}
https://golang.org/pkg/os#UserCacheDir
https://golang.org/pkg/os#UserConfigDir
I have written some Golang code which works when tested on my local machine. When I deploy this as a Google Cloud function it fails because it cannot open a template file. The line of code failing is:
t, err := template.New("list.gohtml").ParseFiles("list.gohtml")
After this call err is set to open list.gohtml: no such file or directory
The file is in the same directory as the go source file and is not listed in .gcloudignore or .gitignore. The gcloud functions documentation says all files in the directory will be uploaded unless listed in one of those ignore files and if I run gcloud meta list-files-for-upload then the file list.gohtml is included in the list displayed.
Is there some magic folder layout to make this work, or an option to the gcloud functions deploy command?
Based on #DazWilkin's reply, I now call the function below at the start of the serving function.
Rather than hardwiring the path into the template file names (which would make it fail when tested locally) this simply checks for the presence of the Gcloud source file directory below the current one, and if present, makes it the working directory, so file resolution will now happen exactly as it does when tested locally.
import "os"
const gcloudFuncSourceDir = "serverless_function_source_code"
func fixDir() {
fileInfo, err := os.Stat(gcloudFuncSourceDir)
if err == nil && fileInfo.IsDir() {
_ = os.Chdir(gcloudFuncSourceDir)
}
}
I created a Function that enumerates the uploaded files:
func L(w http.ResponseWriter, r *http.Request) {
var files []string
err := filepath.Walk(".", func(path string, info os.FileInfo, err error) error {
files = append(files, path)
return nil
})
if err != nil {
panic(err)
}
for _, file := range files {
fmt.Fprintln(w, file)
}
}
It output:
.
go.mod
go.sum
main.go
serverless_function_source_code
serverless_function_source_code/f.go
serverless_function_source_code/go.mod
serverless_function_source_code/test.tmpl
Rewriting the template-using function:
tmpl, err := template.New("test.tmpl").ParseFiles("serverless_function_source_code/test.tmpl")
Works!
However, this is undocumented and a hack :-(
The alternative is to embed the templates as strings within Golang files.
I recommend submitting a feature request on Google's Issue Tracker for Cloud Functions
If there's no go.mod file (in other words, if modules aren't enabled), then the behavior changes and the files are under src/<package-name>/, not under serverless_function_source_code/. The main.go file used to drive the cloud function resides in src/serverless_function_app/main/main.go.
For example:
I have a Go application with the following two files:
listfiles/file_system.go
listfiles/tmpl/letter.html
I deployed it with these commands:
cd listfiles
gcloud functions deploy ListFiles --runtime go113 --trigger-http --allow-unauthenticated
The result was that the current directory was set to /srv. Underneath /srv was a typical GOROOT tree with src/ and pkg/ directories. For example:
.googlebuild
.googlebuild/source-code.tar.gz
pkg
pkg/linux_amd64
pkg/linux_amd64/github.com
pkg/linux_amd64/github.com/GoogleCloudPlatform
pkg/linux_amd64/github.com/GoogleCloudPlatform/functions-framework-go
pkg/linux_amd64/github.com/GoogleCloudPlatform/functions-framework-go/funcframework.a
src
src/cloud.google.com
src/cloud.google.com/go
src/cloud.google.com/go/.git
src/cloud.google.com/go/.git/HEAD
src/cloud.google.com/go/.git/branches
[... snip...]
src/go.uber.org/zap/zaptest/testingt_test.go
src/go.uber.org/zap/zaptest/timeout.go
src/go.uber.org/zap/zaptest/timeout_test.go
src/go.uber.org/zap/zaptest/writer.go
src/go.uber.org/zap/zaptest/writer_test.go
src/listfiles
src/listfiles/file_system.go
src/listfiles/tmpl
src/listfiles/tmpl/letter.html
src/serverless_function_app
src/serverless_function_app/main
src/serverless_function_app/main/main.go
As mentioned in GoDocs, os.Create() creates a file in specific path.
os.Create("fonts/foo/font.eot")
But when fonts or foo doesn't exists, it returns panic: open fonts/foo/font.eot: The system cannot find the path specified.
So i used os.MkdirAll() to create nested directory. But there are many other problems with this function.
path := "fonts/foo/font.eot"
// this line create a directory named (font.eot) !
os.MkdirAll(path, os.ModePerm)
Is there any better way to create a file in nested directories?
The standard way is something like this:
func create(p string) (*os.File, error) {
if err := os.MkdirAll(filepath.Dir(p), 0770); err != nil {
return nil, err
}
return os.Create(p)
}
A few notes:
os.Create does not panic as stated in the question.
Create a directory from the directory part of the file path, not the full path.
I've some command line program which I need to read files from parent folder, I mean
-parentDir
-- myproject
--- cmd
----main.go
--otherdir
-file.json
As you can see otherdir is like sibling to myproject and I need from my main.go read the file.json
what I've tried is like following
func visit(path string, f os.FileInfo, err error) error {
fmt.Printf("Visited: %s\n", path)
return nil
}
func main() {
flag.Parse()
root := flag.Arg(0)
err := filepath.Walk(root, visit)
fmt.Printf("filepath.Walk() returned %v\n", err)
}
I've also try to provide args(-1) which doesnt help...
Any idea how from command line program I can read some files that on level up from my executable ?
I've also tried with
import "github.com/mitchellh/go-homedir"
func Path(path string) (error, string) {
home, err := homedir.Dir()
}
this give the root directory which doesnt help either...
It doesn't matter where the binary is, it matters what the working directory is (the directory you're in when you execute the program). All relative paths will be relative to the current working directory. So, if you're executing from myproject, you'd use something like ../ as the root path to Walk.
That said, I would highly recommend you make the path configurable, rather than assuming the binary will always be executed from some particular location within the source tree.
I have the following structure:
project/
docs/
index.html
root.html
I'm trying to iterate through this project structure so that I can read the contents of each file to process them. So I want to say "search through the directory project", then it will search through all the files, and only the first level of directories and their files, so if there was another directory with a file inside of docs/, it would ignore it.
Currently, I've tried to accomplish this with the "path/filepath" library:
func traverse(path string, file os.FileInfo, err error) error {
if file, err := os.Open(file.Name()); err == nil {
defer file.Close()
if fileStat, err := file.Stat(); err == nil {
switch mode := fileStat.Mode(); {
case mode.IsDir():
fmt.Println("it be a directory! lets traverse", file.Name())
filepath.Walk(file.Name(), traverse)
case mode.IsRegular():
fmt.Println("the thingy ", file.Name(), " is a file")
}
} else {
return errors.New("failed to check status")
}
}
return errors.New("failed 2 open file/dir?")
}
Then I call it from here:
if err := filepath.Walk("project/", traverse); err != nil {
setupErr("%s", err)
}
Note that I run this executable relative to my test directory, so it's finding the directory okay. My problem is actually when I run it, I get the following:
it be a directory! lets traverse project
it be a directory! lets traverse project
# ^ printed about 20 more times ^
failed 2 open file/dir?
I think my recursion is a little off, and it's not changing into the directory perhaps? Any ideas, if you need any more information just ask and I'll update.
First, it looks like what you want do to contradicts with the code you have. You wrote:
So I want to say "search through the directory project", then it will search through all the files, and only the first level of directories and their files, so if there was another directory with a file inside of docs/, it would ignore it.
Does it mean that you want to iterate only two levels of directories (current and one below) and ignore the rest?
If so then you do not need a recursion, just a simple loop that executes search function over the files within the current directory and for every its subdirectory.
The code that you have walks over the filesystem directory subtree.
Basically, filepath.Walk that you use should do it for you. So you either implement recursive walking or use Walk, but not both.
Second, the recursion is implemented incorrectly in your code. It missing iterating over the directories.
So the code that prints the file names in the current directory and its subdirectories (but not further) is:
package main
import (
"fmt"
"io/ioutil"
)
func main() {
items, _ := ioutil.ReadDir(".")
for _, item := range items {
if item.IsDir() {
subitems, _ := ioutil.ReadDir(item.Name())
for _, subitem := range subitems {
if !subitem.IsDir() {
// handle file there
fmt.Println(item.Name() + "/" + subitem.Name())
}
}
} else {
// handle file there
fmt.Println(item.Name())
}
}
}
Walk walks the file tree rooted at root, calling walkFn for each file
or directory in the tree, including root. All errors that arise
visiting files and directories are filtered by walkFn. The files are
walked in lexical order, which makes the output deterministic but
means that for very large directories Walk can be inefficient. Walk
does not follow symbolic links.