From a4038b5fdae2a9cc91b58e745043456486524133 Mon Sep 17 00:00:00 2001
From: schneems <richard.schneeman@gmail.com>
Date: Mon, 18 Jan 2016 09:40:08 -0600
Subject: [PATCH] [close #221] Favor charset from processor

If a charset is returned by a pre-processor use it instead of automatically assigning one. This is needed so that non-gzippable assets will not accidentally be assigned a charset and gzipped later.

We're also removing a hash allocation needed for `Hash.merge!`.
---
 guides/extending_sprockets.md | 9 +++++++++
 lib/sprockets/loader.rb       | 9 ++++-----
 2 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/guides/extending_sprockets.md b/guides/extending_sprockets.md
index d1afae907..ada54c17e 100644
--- a/guides/extending_sprockets.md
+++ b/guides/extending_sprockets.md
@@ -118,6 +118,15 @@ Sprockets expects an array of hashes for this map. Each hash must have a `:sourc
 return {data: data, map: [{ source: "original.coffee", # ... }]}
 ```
 
+- charset: This key contains the mime charset for an asset
+
+A charset is the encoding for text based assets. If you do not specify a charset then one will be automatically assigned by sprockets based on the encoding type of the contents returned in the `:data` key. Normally you want that, the only time you don't want that is if you're working with binary data, or data you don't want to be compressed. If sprockets sees a `charset` then it will think that the contents of the file are text and can be compressed via GZIP. You can avoid this by setting the field manually
+
+```ruby
+return { data: data, charset: nil }
+```
+
+
 WIP the format of the source map may be subject to change before 4.0 is released. Currently it takes a `:original` and `:generated` key which each hold an array of line and column numbers. Line numbers are 1 indexed column numbers are 0 indexed. The first character of a file will always be `[1,0]`.
 
 - WIP - other metadata keys
diff --git a/lib/sprockets/loader.rb b/lib/sprockets/loader.rb
index 0a641bdb1..1eb2fb028 100644
--- a/lib/sprockets/loader.rb
+++ b/lib/sprockets/loader.rb
@@ -160,11 +160,10 @@ def load_from_unloaded(unloaded)
           })
           validate_processor_result!(result)
           source = result.delete(:data)
-          metadata = result.merge!(
-            charset: source.encoding.name.downcase,
-            digest: digest(source),
-            length: source.bytesize
-          )
+          metadata = result
+          metadata[:charset] = source.encoding.name.downcase unless metadata.key?(:charset)
+          metadata[:digest]  = digest(source)
+          metadata[:length]  = source.bytesize
         else
           dependencies << build_file_digest_uri(unloaded.filename)
           metadata = {