#6145 - Introduce deploy script for websites

modules/adblockplus/files/web/static/deploy_script.py

Index: modules/adblockplus/files/web/static/deploy_script.py
===================================================================
new file mode 100644
--- /dev/null
+++ b/modules/adblockplus/files/web/static/deploy_script.py
@@ -0,0 +1,121 @@
+#!/usr/bin/env python
+
+import argparse
+from filecmp import dircmp
+import hashlib
+import os
+import sys
+import shutil
+import tarfile
+import tempfile
+import urllib
+import urllib2
+
+_tmp_dir = tempfile.mkdtemp()
+
+
+def download(url):
+    file_name = url.split('/')[-1]
+    abs_file_name = os.path.join(_tmp_dir, file_name)
+    print 'Downloading: ' + file_name
+    try:
+        filename, _ = urllib.urlretrieve(url, abs_file_name)

[Vasily Kuznetsov, 2018/05/22 16:38:37]

If you're throwing away `filename`, you can just ignore the return value of
`urlretrieve`. Or you can return `filename` from this function, since it now
contains the file where the data was saved.

[f.lopez, 2018/06/06 00:05:03]

Acknowledged.

+        return abs_file_name
+    except urllib2.HTTPError as e:
+        if e.code == 404:
+            sys.exit('File not found on remote source')
+    except Exception as e:
+        sys.exit(e)
+
+
+def calculate_md5(file):
+    with open(file) as f:
+        data = f.read()
+        md5_result = hashlib.md5(data).hexdigest()
+    return md5_result.strip()
+
+
+def read_md5(file):
+    with open(file) as f:
+        md5_result = f.readline()
+    return md5_result.strip()
+
+
+def untar(tar_file):
+    if tarfile.is_tarfile(tar_file):
+        with tarfile.open(tar_file, 'r:gz') as tar:
+            tar.extractall(_tmp_dir)
+
+
+def remove_tree(to_remove):
+    if os.path.exists(to_remove):
+        if os.path.isdir(to_remove):
+            shutil.rmtree(to_remove)
+        else:
+            os.remove(to_remove)
+
+
+def deploy_files(dcmp):
+    for name in dcmp.diff_files:
+        copytree(dcmp.right, dcmp.left)
+    for name in dcmp.left_only:
+        remove_tree(dcmp.left + "/" + name)

[Vasily Kuznetsov, 2018/05/23 12:25:43]

It's more portable (and generally a better practice) to use `os.path.join` for
this.

[f.lopez, 2018/06/06 00:05:03]

Acknowledged.

+    for name in dcmp.right_only:
+        copytree(dcmp.right, dcmp.left)
+    for sub_dcmp in dcmp.subdirs.values():
+        deploy_files(sub_dcmp)
+
+
+def copytree(src, dst):
+    if not os.path.exists(dst):
+        os.makedirs(dst)
+        shutil.copystat(src, dst)
+    lst = os.listdir(src)
+    for item in lst:
+        s = os.path.join(src, item)
+        d = os.path.join(dst, item)
+        if os.path.isdir(s):
+            copytree(s, d)
+        else:
+            shutil.copy2(s, d)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(
+        description="""Fetch a compressed archive in the form of $HASH.tar.gz
+                    and deploy it to /var/www/$WEBSITE folder""",
+        epilog="""--hash must be provided in order to fetch the files,
+               expected files to be fetched are $HASH.tar.gz and $HASH.md5 in
+               order to compare the hashes.
+               --source must be an URL, e.g.
+               https://helpcenter.eyeofiles.com""",
+    )
+    parser.add_argument('--hash', action='store', type=str,
+                        nargs='?', required=True,
+                        help='Hash of the commit to deploy')
+    parser.add_argument('--source', action='store', type=str,
+                        required=True, nargs='?',

[Vasily Kuznetsov, 2018/05/22 16:38:37]

Won't nargs='?' make the argument (of --source) optional? And then what happens
if it's not provided (if --source is given without any argument)?

[f.lopez, 2018/06/06 00:05:03]

Acknowledged.

+                        help='The source where files will be downloaded')
+    parser.add_argument('--website', action='store', type=str, nargs='?',
+                        help='The name of the website [e.g. help.eyeo.com]')
+    args = parser.parse_args()
+    hash = args.hash
+    source = args.source
+    url_file = '{0}/{1}.tar.gz'.format(source, hash)
+    url_md5 = '{0}/{1}.md5'.format(source, hash)
+    down_file = download(url_file)
+    down_md5 = download(url_md5)
+    try:
+        if calculate_md5(down_file) == read_md5(down_md5):
+            untar(down_file)
+            hash_directory = os.path.join(_tmp_dir, hash)
+            destination = '/var/www/' + args.website

[Vasily Kuznetsov, 2018/05/23 12:25:43]

Maybe also `os.path.join`?

Also, do you think it's ok to hardcode '/var/www/'? Perhaps better to make it an
option (maybe with this default)? It's a genuine question -- if this thing is
never expected to change, perhaps it's ok.

[f.lopez, 2018/06/06 00:05:03]

This is not expected to change, since it is the default value we look for in the
infrastructure part

+            dcmp = dircmp(destination, hash_directory)
+            print "Deploying files"

[Vasily Kuznetsov, 2018/05/23 12:25:44]

Remember the quotes rule ;P

[f.lopez, 2018/06/06 00:05:03]

Oh my god...

+            deploy_files(dcmp)
+        else:
+            sys.exit("Hashes don't match")
+    except Exception as e:
+        sys.exit(e)
+    finally:
+        shutil.rmtree(_tmp_dir)

[Vasily Kuznetsov, 2018/05/23 12:25:44]

The temporary directory won't be deleted if download() fails in lines 106 and
107 or if argument parsing throws an exception. I would propose to:
- move the creation of temporary directory into the `if __name__ == '__main__'
block,
- create it only before it's actually needed (before download()),
- pass it into download() as a parameter,
- move the download() calls into the try-finally block.
This way the temporary directory will always be deleted.

The side effect of this will be that download() will be called in the same
try-except block and you can do all exception handling only in lines 118-119...
and reduce duplication.

[f.lopez, 2018/06/06 00:05:03]

ok I like it, thanks