Support per-repo http_proxy (#101)

This adds a feature with which http_proxy could be set on a per-repo basis.

This is useful if pacoloco runs on a server where most of the repos it serve are easy to access, but some others are not, e.g.:

This is my config:

```
repos:
  archlinux:
    urls:
      - http://mirrors.wsyu.edu.cn/archlinux
      - http://mirrors.163.com/archlinux
      - https://mirrors.aliyun.com/archlinux
      - http://mirrors.tuna.tsinghua.edu.cn/archlinux
  archlinuxarm:
    urls:
      - http://mirrors.163.com/archlinuxarm
      - https://mirrors.aliyun.com/archlinuxarm
  archlinuxcn_x86_64:
    urls:
      - http://mirrors.wsyu.edu.cn/archlinucn
      - http://mirrors.163.com/archlinucn
      - https://mirrors.aliyun.com/archlinuxcn
  archlinuxcn_aarch64:
    urls:
      - http://mirrors.wsyu.edu.cn/archlinucn
      - http://mirrors.163.com/archlinucn
      - https://mirrors.aliyun.com/archlinuxcn
  7Ji_x86_64:
    http_proxy: http://proxy.lan:1085
    url: https://github.com/7Ji/archrepo/releases/download
  7Ji_aarch64:
    http_proxy: http://proxy.lan:1085
    url: https://github.com/7Ji/archrepo/releases/download
```

As the config shows, there're mirrors for `archlinux`, `archlinuxcn` and `archlinuxarm` in China so I don't want to set a global http_proxy to waste my proxy bandwidth, but repo 7Ji uses Github releases as storage and it is hard to access without a proxy in China, so I want to use proxy for and only for repo 7Ji.

README.md

@@ -100,6 +100,7 @@ repos:
   quarry:
     url: http://pkgbuild.com/~anatolik/quarry/x86_64
   sublime:
+    http_proxy: http://bar.company.com:8989 # Proxy could be enabled per-repo, shadowing the global `http_proxy` (see below)
     url: https://download.sublimetext.com/arch/stable/x86_64
   archlinux-reflector:
     mirrorlist: /etc/pacman.d/reflector_mirrorlist # Be careful! Check that pacoloco URL is NOT included in that file!
@@ -116,7 +117,7 @@ prefetch: # optional section, add it if you want to enable prefetching
 * `purge_files_after` specifies inactivity duration (in seconds) after which the file should be removed from the cache. This functionality uses unix "AccessTime" field to find out inactive files. Default value is `0` that means never run the purging.
 * `port` is the server port.
 * `download_timeout` is a timeout (in seconds) for internet->cache downloads. If a remote server gets slow and file download takes longer than this will be terminated. Default value is `0` that means no timeout.
-* `repos` is a list of repositories to mirror. Each repo needs `name` and url of its Arch mirrors. Note that url can be specified either with `url` or `urls` properties, one and only one can be used for each repo configuration.
+* `repos` is a list of repositories to mirror. Each repo needs `name` and url of its Arch mirrors. Note that url can be specified either with `url` or `urls` properties, one and only one can be used for each repo configuration. Each repo could have its own `http_proxy`, which would shadow the global `http_proxy` (see below).
 * `http_proxy` is only to be used if you have pacoloco running behind a proxy
 * `user_agent` user agent used to fetch the files from repositories. Default value is `Pacoloco/1.2`.
 * The `prefetch` section allows to enable packages prefetching. Comment it out to disable it.

config.go

@@ -22,6 +22,7 @@ type Repo struct {
 	URL                  string    `yaml:"url"`
 	URLs                 []string  `yaml:"urls"`
 	Mirrorlist           string    `yaml:"mirrorlist"`
+	HttpProxy            string    `yaml:"http_proxy"`
 	LastMirrorlistCheck  time.Time `yaml:"-"`
 	LastModificationTime time.Time `yaml:"-"`
 }

downloader.go

@@ -6,6 +6,7 @@ import (
 	"io"
 	"log"
 	"net/http"
+	"net/url"
 	"os"
 	"path"
 	"path/filepath"
@@ -62,8 +63,15 @@ func (d *Downloader) download() error {
 		return fmt.Errorf("repo %v has no urls", d.repoName)
 	}
 
+	var proxyURL *url.URL
+	if d.repo.HttpProxy != "" {
+		proxyURL, _ = url.Parse(d.repo.HttpProxy)
+	} else {
+		proxyURL = nil
+	}
+
 	for _, u := range d.repo.getUrls() {
-		err := d.downloadFromUpstream(u)
+		err := d.downloadFromUpstream(u, proxyURL)
 		if err != nil {
 			log.Printf("unable to download file %v: %v", d.key, err)
 			continue // try next mirror
@@ -73,7 +81,7 @@ func (d *Downloader) download() error {
 	return fmt.Errorf("unable to download file %v", d.key)
 }
 
-func (d *Downloader) downloadFromUpstream(repoURL string) error {
+func (d *Downloader) downloadFromUpstream(repoURL string, proxyURL *url.URL) error {
 	upstreamURL := repoURL + d.urlPath
 
 	var req *http.Request
@@ -101,7 +109,16 @@ func (d *Downloader) downloadFromUpstream(repoURL string) error {
 
 	log.Printf("downloading %v", upstreamURL)
 
-	resp, err := http.DefaultClient.Do(req)
+	var client *http.Client
+	if proxyURL != nil {
+		proxy := http.ProxyURL(proxyURL)
+		transport := &http.Transport{Proxy: proxy}
+		client = &http.Client{Transport: transport}
+	} else {
+		client = http.DefaultClient
+	}
+
+	resp, err := client.Do(req)
 	if err != nil {
 		return err
 	}

pacoloco.yaml.sample

@@ -13,6 +13,7 @@ repos:
     mirrorlist: /etc/pacman.d/mirrorlist ## Be careful! Check that pacoloco URL is NOT included in that file!
 ## Local/3rd party repos can be added following the below example:
 #  quarry:
+#    http_proxy: http://bar.company.com:8989 ## Proxy could be enabled per-repo, shadowing the global `http_proxy` (see below)
 #    url: http://pkgbuild.com/~anatolik/quarry/x86_64
 
 # prefetch: ## optional section, add it if you want to enable prefetching