atob, btoa: handle latin-1 correctly

1 files changed, 36 insertions, 12 deletions

diff --git a/src/js/base64.nim b/src/js/base64.nim
index 9b61f604..1ff81cc3 100644
--- a/src/js/base64.nim
+++ b/src/js/base64.nim
@@ -1,24 +1,48 @@
-import unicode
 import std/base64
 
 import js/domexception
 import utils/opt
 
+# atob and btoa convert Latin-1 to base64 and vice versa. (And throw on
+# anything above latin-1.)
+# We could do this quite efficiently if we had an API for the QuickJS string
+# internal representation. Unfortunately we do not, so we do the following:
+# * atob: decode, convert latin-1 to utf-8, pass to qjs, where it is then
+#   converted to latin-1 again.
+# * btoa: qjs converts its string (either utf-16 or latin-1) to utf-8,
+#   we convert this to latin-1 (or throw), then encode.
+# That is two conversions more than needed (i.e. 0) for each step. We should
+# really write an API for handling QJS strings sometime...
+
 proc atob*(data: string): DOMResult[string] =
   try:
-    return ok(base64.decode(data))
+    let ds = base64.decode(data)
+    var s = newStringOfCap(ds.len)
+    for c in ds:
+      if uint8(c) <= 0x7F:
+        s &= c
+      else: # latin-1
+        s &= char((uint8(c) shr 6) or 0xC0)
+        s &= char((uint8(c) and 0x3F) or 0x80)
+    return ok(s)
   except ValueError:
     return err(newDOMException("Invalid character in string",
       "InvalidCharacterError"))
 
 proc btoa*(data: string): DOMResult[string] =
-  # For some incomprehensible reason, this must return a DOMException
-  # for strings with chars not in latin-1.
-  #TODO this could be a lot more efficient if we just directly passed
-  # is_wide_char from QJS strings.
-  for r in data.runes:
-    if int32(r) <= 0xFF:
-      continue
-    return err(newDOMException("Invalid character in string",
-      "InvalidCharacterError"))
-  return ok(base64.encode(data))
+  var s = newStringOfCap(data.len)
+  var i = 0
+  while i < data.len:
+    let c = data[i]
+    let n = uint8(c)
+    if n <= 0x7F: # ascii
+      s &= c
+      inc i
+    elif n <= 0xC3: # latin-1
+      inc i
+      s &= char((n shl 6) or (uint8(data[i]) and 0x3F))
+      inc i
+    else:
+      return err(newDOMException("Invalid character in string",
+        "InvalidCharacterError"))
+  return ok(base64.encode(s))