Remove an allocation when decoding base64

email.d

@@ -9,6 +9,9 @@ pragma(lib, "curl");
 import std.base64;
 import std.string;
 import std.range;
+import std.utf;
+import std.array;
+import std.algorithm.iteration;
 
 import arsd.characterencodings;
 
@@ -866,18 +869,10 @@ class IncomingEmailMessage {
 				break;
 				case "base64":
 					if(textMessageBody.length) {
-						// alas, phobos' base64 decoder cannot accept ranges, so we have to allocate here
+						textMessageBody = textMessageBody.decodeBase64Mime.convertToUtf8Lossy(charset);
-						char[] mmb;
-						mmb.reserve(textMessageBody.length);
-						foreach (char ch; textMessageBody) if (ch > ' ' && ch < 127) mmb ~= ch;
-						textMessageBody = convertToUtf8Lossy(Base64.decode(mmb), charset);
 					}
 					if(htmlMessageBody.length) {
-						// alas, phobos' base64 decoder cannot accept ranges, so we have to allocate here
+						htmlMessageBody = htmlMessageBody.decodeBase64Mime.convertToUtf8Lossy(charset);
-						char[] mmb;
-						mmb.reserve(htmlMessageBody.length);
-						foreach (char ch; htmlMessageBody) if (ch > ' ' && ch < 127) mmb ~= ch;
-						htmlMessageBody = convertToUtf8Lossy(Base64.decode(mmb), charset);
 					}
 
 				break;
@@ -1172,6 +1167,20 @@ string encodeBase64Mime(const(ubyte[]) content, string LINESEP = "\r\n") {
 	return cast(immutable(char[]))content.chunks(SOURCE_CHUNK_LENGTH).base64encode.join(LINESEP);
 }
 
+/// Base64 range decoder UFCS helper.
+alias base64decode = Base64.decoder;
+
+/// Base64 decoder, ignoring linebreaks which are mandated by RFC2045
+immutable(ubyte[]) decodeBase64Mime(string encodedPart) {
+	return cast(immutable(ubyte[])) encodedPart
+		.byChar // prevent Autodecoding, which will break Base64 decoder. Since its base64, it's guarenteed to be 7bit ascii
+		.filter!((c) => c != '\r')
+		.splitter!((c) => c == '\n')
+		.joiner
+		.base64decode
+		.array;
+}
+
 /+
 void main() {
 	import std.file;