Skip over function handles in v7.3 format

This is similar to JuliaIO#129 which returns `missing` for unhandled data
types in HDF5 Datasets; function handles are stored as HDF5 groups, so
a different explicit check is required to skip over them.

src/MAT_HDF5.jl

@@ -229,6 +229,9 @@ function m_read(g::HDF5Group)
                 data = str2eltype_matlab[mattype][]
             end
             return SparseMatrixCSC(convert(Int, HDF5.a_read(g, sparse_attr_matlab)), length(jc)-1, jc, ir, data)
+        elseif mattype == "function_handle"
+            @warn "MATLAB $mattype values are currently not supported"
+            return missing
         else
             error("Cannot read from a non-struct group, type was $mattype")
         end

test/read.jl

@@ -213,9 +213,16 @@ let objtestfile = "figure.fig"
     @test vars["hgS_070000"]["type"] == "figure"
 end
 
-# test reading file containing Matlab table and datetime objects
+# test reading file containing Matlab function handle, table, and datetime objects
 # since we don't support these objects, just make sure that there are no errors
 # reading the file and that the variables are there and replaced with `missing`
+let objtestfile = "function_handles.mat"
+    vars = matread(joinpath(dirname(@__FILE__), "v7.3", objtestfile))
+    @test "sin" in keys(vars)
+    @test ismissing(vars["sin"])
+    @test "anonymous" in keys(vars)
+    @test ismissing(vars["anonymous"])
+end
 let objtestfile = "struct_table_datetime.mat"
     vars = matread(joinpath(dirname(@__FILE__), "v7.3", objtestfile))["s"]
     @test "testTable" in keys(vars)