Save options, test cases for missing labels

Co-authored-by: Ulf Wiger <ulf@wiger.net>
Reviewed-on: #51

README.md

@@ -58,6 +58,24 @@ Labels correspond to (existing) atoms in Erlang.
 Decoding of a label results in a call to `binary_to_existing_atom/2`, so will
 fail if the corresponding atom does not already exist.
 
+This behavior can be modified using the option `#{missing_labels => fail | create | convert}`,
+where `fail` is the default, as described above, `convert` means that missing atoms are
+converted to binaries, and `create` means that the atom is created dynamically.
+
+The option can be passed e.g.:
+```erlang
+gmser_dyn:deserialize(Binary, set_opts(#{missing_labels => convert}))
+```
+
+or
+```erlang
+gmser_dyn:deserialize(Binary, set_opts(#{missing_labels => convert}, Types))
+```
+
+By calling `gmser_dyn:register_types/1`, after having added options to the type map,
+the options can be made to take effect automatically.
+
+
 It's possible to cache labels for more compact encoding.
 Note that when caching labels, the same cache mapping needs to be used on the
 decoder side.

src/gmser_dyn.erl

@@ -17,6 +17,9 @@
 %% register a type schema, inspect existing schema
 -export([ register_types/1
         , registered_types/0
+        , get_opts/1
+        , set_opts/1
+        , set_opts/2
         , types_from_list/1
         , revert_to_default_types/0
         , dynamic_types/0 ]).
@@ -121,6 +124,7 @@ dynamic_types() ->
             , id     => id
             , label  => label
             }
+     , options => #{}
        }.
 
 vsn(Types) ->
@@ -326,13 +330,13 @@ encode_from_template(Type, List, E, Vsn, Types) when is_list(Type), is_list(List
     encode_fields(Type, List, E, Vsn, Types);
 encode_from_template(label, V, E, _, Types) ->
     assert_type(is_atom(V), label, V),
-    emit(E, label, Types,
-         case find_cached_label(V, Types) of
-             error ->
-                 encode_basic(label, V, E, Types);
-             {ok, Code} when is_integer(Code) ->
-                 [encode_basic(int, Code)]
-         end);
+    case find_cached_label(V, Types) of
+        error ->
+            encode_basic(label, V, E, Types);
+        {ok, Code} when is_integer(Code) ->
+            emit(E, label, Types,
+                 [encode_basic(int, Code)])
+    end;
 encode_from_template(Type, V, E, _, Types) when Type == id
                                                 ; Type == binary
                                                 ; Type == bool
@@ -399,15 +403,29 @@ emit(0, _, _, Enc) ->
 emit_code(Tag, #{rev := Tags}) ->
     encode_basic(int, maps:get(Tag, Tags)).
 
-decode_basic(Type, [Tag,V], #{codes := Codes}) ->
+decode_basic(Type, [Tag,V], #{codes := Codes} = Types) ->
     case decode_basic(int, Tag) of
         Code when map_get(Code, Codes) == Type ->
-            decode_basic(Type, V);
+            decode_basic_(Type, V, Types);
         _ ->
             error(illegal)
     end;
-decode_basic(Type, V, _) ->
-    decode_basic(Type, V).
+decode_basic(Type, V, Types) ->
+    decode_basic_(Type, V, Types).
+
+decode_basic_(label, Fld, #{options := #{missing_labels := Opt}}) ->
+    Bin = decode_basic(binary, Fld),
+    case Opt of
+        create  -> binary_to_atom(Bin, utf8);
+        fail    -> binary_to_existing_atom(Bin, utf8);
+        convert ->
+            try binary_to_existing_atom(Bin, utf8)
+            catch
+                error:_ -> Bin
+            end
+    end;
+decode_basic_(Type, Fld, _) ->
+    decode_basic(Type, Fld).
 
 decode_basic(label, Fld) ->
     binary_to_existing_atom(decode_basic(binary, Fld), utf8);
@@ -443,11 +461,14 @@ register_types(Types) when is_map(Types) ->
     Rev   = rev_codes(Codes),
     Templates = maps:get(templates, Types, #{}),
     Labels    = maps:get(labels, Types, #{}),
-    #{codes := Codes0, rev := Rev0, labels := Labels0, templates := Templates0} =
+    Options   = maps:get(options, Types, #{}),
+    #{codes := Codes0, rev := Rev0, labels := Labels0,
+      templates := Templates0, options := Options0} =
         dynamic_types(),
     Merged = #{ codes => maps:merge(Codes0, Codes)
               , rev   => maps:merge(Rev0, Rev)
               , templates => maps:merge(Templates0, Templates)
+              , options => maps:merge(Options0, Options)
               , labels    => maps:merge(Labels0, Labels) },
     assert_sizes(Merged),
     assert_mappings(Merged),
@@ -473,6 +494,15 @@ register_type(Code, Tag, Template) when is_integer(Code), Code >= 0 ->
         {_, true} -> error(tag_exists)
     end.
 
+set_opts(Opts) ->
+    set_opts(Opts, registered_types()).
+
+set_opts(Opts, Types) ->
+    Types#{options => Opts}.
+
+get_opts(#{options := Opts}) ->
+    Opts.
+
 cache_label(Code, Label) when is_integer(Code), Code >= 0, is_atom(Label) ->
     #{labels := Lbls, rev_labels := RevLbls} = Types = registered_types(),
     case {is_map_key(Label, Lbls), is_map_key(Code, RevLbls)} of
@@ -623,6 +653,7 @@ user_types_test_() ->
      , ?_test(t_reg_template_vsnd_fun())
      , ?_test(t_reg_label_cache())
      , ?_test(t_reg_label_cache2())
+     , ?_test(t_reg_options())
      ]}.
 
 dynamic_types_test_() ->
@@ -630,6 +661,7 @@ dynamic_types_test_() ->
     , ?_test(t_typed_map())
     , ?_test(t_alts())
     , ?_test(t_anyints())
+    , ?_test(t_missing_labels())
     ].
 
 t_round_trip(T) ->
@@ -723,6 +755,13 @@ t_reg_label_cache2() ->
     [<<0>>,<<1>>,[<<3,235>>,[[<<49>>],[<<49>>]]]] = Enc,
     _Tup = gmser_dyn:decode(Enc).
 
+t_reg_options() ->
+    register_types(set_opts(#{missing_labels => convert})),
+    [Dyn,Vsn,[Am,<<"random">>]] = gmser_dyn:encode(random),
+    EncNewAm = [Dyn,Vsn,[Am,<<"foo12345">>]],
+    <<"foo12345">> = gmser_dyn:decode(EncNewAm),
+    ok.
+
 t_typed_map() ->
     Term = #{a => 13, {key,1} => [a]},
     Enc = encode_typed(#{items => [{a,int},{{key,1},[label]}]}, Term),
@@ -738,4 +777,13 @@ t_anyints() ->
     t_round_trip_typed(anyint, 5),
     ok.
 
+t_missing_labels() ->
+    [Dyn,Vsn,[Am,<<"random">>]] = gmser_dyn:encode(random),
+    EncNewAm = [Dyn,Vsn,[Am,<<"flurbee">>]],
+    ?assertError(badarg, gmser_dyn:decode(EncNewAm)),
+    ?assertError(badarg, gmser_dyn:decode(EncNewAm, set_opts(#{missing_labels => fail}))),
+    <<"flurbee">> = gmser_dyn:decode(EncNewAm, set_opts(#{missing_labels => convert})),
+    true = is_atom(gmser_dyn:decode(EncNewAm, set_opts(#{missing_labels => create}))),
+    ok.
+
 -endif.