hugo/tpl/template_resources.go

256 lines
6.9 KiB
Go
Raw Normal View History

// Copyright 2016 The Hugo Authors. All rights reserved.
//
2015-11-24 03:16:36 +00:00
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
2015-11-24 03:16:36 +00:00
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package tpl
import (
"bytes"
"encoding/csv"
"encoding/json"
"errors"
"io/ioutil"
"net/http"
"net/url"
"path/filepath"
"strings"
"sync"
"time"
"github.com/spf13/afero"
"github.com/spf13/hugo/helpers"
"github.com/spf13/hugo/hugofs"
jww "github.com/spf13/jwalterweatherman"
"github.com/spf13/viper"
)
var (
remoteURLLock = &remoteLock{m: make(map[string]*sync.Mutex)}
resSleep = time.Second * 2 // if JSON decoding failed sleep for n seconds before retrying
resRetries = 1 // number of retries to load the JSON from URL or local file system
)
type remoteLock struct {
sync.RWMutex
m map[string]*sync.Mutex
}
// URLLock locks an URL during download
func (l *remoteLock) URLLock(url string) {
l.Lock()
if _, ok := l.m[url]; !ok {
l.m[url] = &sync.Mutex{}
}
l.Unlock() // call this Unlock before the next lock will be called. NFI why but defer doesn't work.
l.m[url].Lock()
}
// URLUnlock unlocks an URL when the download has been finished. Use only in defer calls.
func (l *remoteLock) URLUnlock(url string) {
l.RLock()
defer l.RUnlock()
if um, ok := l.m[url]; ok {
um.Unlock()
}
}
// getCacheFileID returns the cache ID for a string
func getCacheFileID(id string) string {
return viper.GetString("cacheDir") + url.QueryEscape(id)
}
// resGetCache returns the content for an ID from the file cache or an error
// if the file is not found returns nil,nil
func resGetCache(id string, fs afero.Fs, ignoreCache bool) ([]byte, error) {
if ignoreCache {
return nil, nil
}
fID := getCacheFileID(id)
isExists, err := helpers.Exists(fID, fs)
if err != nil {
return nil, err
}
if !isExists {
return nil, nil
}
return afero.ReadFile(fs, fID)
}
// resWriteCache writes bytes to an ID into the file cache
func resWriteCache(id string, c []byte, fs afero.Fs, ignoreCache bool) error {
if ignoreCache {
return nil
}
fID := getCacheFileID(id)
f, err := fs.Create(fID)
if err != nil {
return errors.New("Error: " + err.Error() + ". Failed to create file: " + fID)
}
defer f.Close()
n, err := f.Write(c)
if n == 0 {
return errors.New("No bytes written to file: " + fID)
}
if err != nil {
return errors.New("Error: " + err.Error() + ". Failed to write to file: " + fID)
}
return nil
}
func resDeleteCache(id string, fs afero.Fs) error {
return fs.Remove(getCacheFileID(id))
}
// resGetRemote loads the content of a remote file. This method is thread safe.
func resGetRemote(url string, fs afero.Fs, hc *http.Client) ([]byte, error) {
c, err := resGetCache(url, fs, viper.GetBool("ignoreCache"))
if c != nil && err == nil {
return c, nil
}
if err != nil {
return nil, err
}
// avoid race condition with locks, block other goroutines if the current url is processing
remoteURLLock.URLLock(url)
defer func() { remoteURLLock.URLUnlock(url) }()
// avoid multiple locks due to calling resGetCache twice
c, err = resGetCache(url, fs, viper.GetBool("ignoreCache"))
if c != nil && err == nil {
return c, nil
}
if err != nil {
return nil, err
}
jww.INFO.Printf("Downloading: %s ...", url)
res, err := hc.Get(url)
if err != nil {
return nil, err
}
c, err = ioutil.ReadAll(res.Body)
res.Body.Close()
if err != nil {
return nil, err
}
err = resWriteCache(url, c, fs, viper.GetBool("ignoreCache"))
if err != nil {
return nil, err
}
jww.INFO.Printf("... and cached to: %s", getCacheFileID(url))
return c, nil
}
// resGetLocal loads the content of a local file
func resGetLocal(url string, fs afero.Fs) ([]byte, error) {
filename := filepath.Join(viper.GetString("workingDir"), url)
if e, err := helpers.Exists(filename, fs); !e {
return nil, err
}
return afero.ReadFile(fs, filename)
}
// resGetResource loads the content of a local or remote file
func resGetResource(url string) ([]byte, error) {
if url == "" {
return nil, nil
}
if strings.Contains(url, "://") {
return resGetRemote(url, hugofs.Source(), http.DefaultClient)
}
return resGetLocal(url, hugofs.Source())
}
// getJSON expects one or n-parts of a URL to a resource which can either be a local or a remote one.
// If you provide multiple parts they will be joined together to the final URL.
// GetJSON returns nil or parsed JSON to use in a short code.
func getJSON(urlParts ...string) interface{} {
var v interface{}
url := strings.Join(urlParts, "")
for i := 0; i <= resRetries; i++ {
c, err := resGetResource(url)
if err != nil {
jww.ERROR.Printf("Failed to get json resource %s with error message %s", url, err)
return nil
}
err = json.Unmarshal(c, &v)
if err != nil {
jww.ERROR.Printf("Cannot read json from resource %s with error message %s", url, err)
jww.ERROR.Printf("Retry #%d for %s and sleeping for %s", i, url, resSleep)
time.Sleep(resSleep)
resDeleteCache(url, hugofs.Source())
continue
}
break
}
return v
}
// parseCSV parses bytes of CSV data into a slice slice string or an error
func parseCSV(c []byte, sep string) ([][]string, error) {
if len(sep) != 1 {
return nil, errors.New("Incorrect length of csv separator: " + sep)
}
b := bytes.NewReader(c)
r := csv.NewReader(b)
rSep := []rune(sep)
r.Comma = rSep[0]
r.FieldsPerRecord = 0
return r.ReadAll()
}
// getCSV expects a data separator and one or n-parts of a URL to a resource which
// can either be a local or a remote one.
// The data separator can be a comma, semi-colon, pipe, etc, but only one character.
// If you provide multiple parts for the URL they will be joined together to the final URL.
// GetCSV returns nil or a slice slice to use in a short code.
func getCSV(sep string, urlParts ...string) [][]string {
var d [][]string
url := strings.Join(urlParts, "")
var clearCacheSleep = func(i int, u string) {
jww.ERROR.Printf("Retry #%d for %s and sleeping for %s", i, url, resSleep)
time.Sleep(resSleep)
resDeleteCache(url, hugofs.Source())
}
for i := 0; i <= resRetries; i++ {
c, err := resGetResource(url)
if err == nil && false == bytes.Contains(c, []byte(sep)) {
err = errors.New("Cannot find separator " + sep + " in CSV.")
}
if err != nil {
jww.ERROR.Printf("Failed to read csv resource %s with error message %s", url, err)
clearCacheSleep(i, url)
continue
}
if d, err = parseCSV(c, sep); err != nil {
jww.ERROR.Printf("Failed to parse csv file %s with error message %s", url, err)
clearCacheSleep(i, url)
continue
}
break
}
return d
}