The current parser assumes that the data fields in a (sub)chunk fills

up the entire chunk size. I have encountered WAV files with chunk padding, that means the next chunk starts only after a few unused padding bytes. Before reading the next subchunk header, you should explicitly seek to the next subchunk start as indicated by the actual chunk size in the chunk header instead on relying on things lining up. I've fixed the issue (dancasimiro#90), and while I was at it I've also fixed the restriction of the format and data chunk order.
CMGeldenhuys · Oct 3, 2020 · 89a4036 · 89a4036
1 parent e182794
commit 89a4036
Showing 1 changed file with 13 additions and 7 deletions.
diff --git a/src/WAV.jl b/src/WAV.jl
@@ -758,20 +758,20 @@ function wavread(io::IO; subrange=(:), format="double")
     sample_rate = Float32(0.0)
     opt = WAVChunk[]
 
-    # Note: This assumes that the format chunk is written in the file before the data chunk. The
-    # specification does not require this assumption, but most real files are written that way.
-
     # Subtract the size of the format field from chunk_size; now it holds the size
     # of all the sub-chunks
     chunk_size -= 4
     # GitHub Issue #18: Check if there is enough data to read another chunk
     subchunk_header_size = 4 + sizeof(UInt32)
     fmt = WAVFormat()
+    data_position = 0
+    data_size = 0
     while chunk_size >= subchunk_header_size
         # Read subchunk ID and size
         subchunk_id = Vector{UInt8}(undef, 4)
         read!(io, subchunk_id)
         subchunk_size = read_le(io, UInt32)
+        nextchunk_start = position(io) + subchunk_size
         if subchunk_size > chunk_size
             chunk_size = 0
             break
@@ -784,15 +784,21 @@ function wavread(io::IO; subrange=(:), format="double")
             nbits = bits_per_sample(fmt)
             push!(opt, WAVChunk(fmt))
         elseif subchunk_id == b"data"
-            if format == "size"
-                return convert(Int, subchunk_size / fmt.block_align), convert(Int, fmt.nchannels)
-            end
-            samples = read_data(io, subchunk_size, fmt, format, make_range(subrange))
+            data_position = position(io)
+            data_size = subchunk_size
         else
             subchunk_data = Vector{UInt8}(undef, subchunk_size)
             read!(io, subchunk_data)
             push!(opt, WAVChunk(Symbol(subchunk_id), subchunk_data))
         end
+        seek(io, nextchunk_start)
+    end
+    if data_size > 0 && data_position > 0
+        seek(io, data_position)
+        if format == "size"
+            return convert(Int, data_size / fmt.block_align), convert(Int, fmt.nchannels)
+        end
+        samples = read_data(io, data_size, fmt, format, make_range(subrange))
     end
     return samples, sample_rate, nbits, opt
 end