%%
%% %CopyrightBegin%
%%
%% Copyright Ericsson AB 2012-2013. All Rights Reserved.
%%
%% The contents of this file are subject to the Erlang Public License,
%% Version 1.1, (the "License"); you may not use this file except in
%% compliance with the License. You should have received a copy of the
%% Erlang Public License along with this software. If not, it can be
%% retrieved online at http://www.erlang.org/.
%%
%% Software distributed under the License is distributed on an "AS IS"
%% basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
%% the License for the specific language governing rights and limitations
%% under the License.
%%
%% %CopyrightEnd%
%%
-module(asn1rtt_real_common).
-export([encode_real/1,decode_real/1,
ber_encode_real/1]).
%%============================================================================
%%
%% Real value, ITU_T X.690 Chapter 8.5
%%============================================================================
%%
%% encode real value
%%============================================================================
ber_encode_real(0) ->
{[],0};
ber_encode_real('PLUS-INFINITY') ->
{[64],1};
ber_encode_real('MINUS-INFINITY') ->
{[65],1};
ber_encode_real(Val) when is_tuple(Val); is_list(Val) ->
encode_real(Val).
%%%%%%%%%%%%%%
%% only base 2 encoding!
%% binary encoding:
%% +------------+ +------------+ +-+-+-+-+---+---+
%% | (tag)9 | | n + p + 1 | |1|S|BB |FF |EE |
%% +------------+ +------------+ +-+-+-+-+---+---+
%%
%% +------------+ +------------+
%% | | | |
%% +------------+ ...+------------+
%% n octets for exponent
%%
%% +------------+ +------------+
%% | | | |
%% +------------+ ...+------------+
%% p octets for pos mantissa
%%
%% S is 0 for positive sign
%% 1 for negative sign
%% BB: encoding base, 00 = 2, (01 = 8, 10 = 16)
%% 01 and 10 not used
%% FF: scale factor 00 = 0 (used in base 2 encoding)
%% EE: encoding of the exponent:
%% 00 - on the following octet
%% 01 - on the 2 following octets
%% 10 - on the 3 following octets
%% 11 - encoding of the length of the two's-complement encoding of
%% exponent on the following octet, and two's-complement
%% encoding of exponent on the other octets.
%%
%% In DER and base 2 encoding the mantissa is encoded as value 0 or
%% bit shifted until it is an odd number. Thus, do this for BER as
%% well.
encode_real(Real) ->
encode_real([], Real).
encode_real(_C, {Mantissa, Base, Exponent}) when Base =:= 2 ->
%% io:format("Mantissa: ~w Base: ~w, Exp: ~w~n",[Man, Base, Exp]),
{Man,ExpAdd} = truncate_zeros(Mantissa), %% DER adjustment
Exp = Exponent + ExpAdd,
OctExp = if Exp >= 0 -> list_to_binary(encode_pos_integer(Exp, []));
true -> list_to_binary(encode_neg_integer(Exp, []))
end,
%% ok = io:format("OctExp: ~w~n",[OctExp]),
SignBit = if Man > 0 -> 0; % bit 7 is pos or neg, no Zeroval
true -> 1
end,
%% ok = io:format("SignBitMask: ~w~n",[SignBitMask]),
SFactor = 0,
OctExpLen = byte_size(OctExp),
if OctExpLen > 255 ->
exit({error,{asn1, {to_big_exp_in_encode_real, OctExpLen}}});
true -> true %% make real assert later..
end,
{LenCode, EOctets} = case OctExpLen of % bit 2,1
1 -> {0, OctExp};
2 -> {1, OctExp};
3 -> {2, OctExp};
_ -> {3, <<OctExpLen, OctExp/binary>>}
end,
BB = 0, %% 00 for base 2
FirstOctet = <<1:1,SignBit:1,BB:2,SFactor:2,LenCode:2>>,
OctMantissa = if Man > 0 -> list_to_binary(real_mininum_octets(Man));
true -> list_to_binary(real_mininum_octets(-(Man))) % signbit keeps track of sign
end,
%% ok = io:format("LenMask: ~w EOctets: ~w~nFirstOctet: ~w OctMantissa: ~w OctExpLen: ~w~n", [LenMask, EOctets, FirstOctet, OctMantissa, OctExpLen]),
Bin = <<FirstOctet/binary, EOctets/binary, OctMantissa/binary>>,
{Bin, size(Bin)};
encode_real(C, {Mantissa,Base,Exponent})
when Base =:= 10, is_integer(Mantissa), is_integer(Exponent) ->
%% always encode as NR3 due to DER on the format
%% mmmm.Eseeee where
%% m := digit
%% s := '-' | '+' | []
%% '+' only allowed in +0
%% e := digit
%% ex: 1234.E-5679
ManStr = integer_to_list(Mantissa),
encode_real_as_string(C,ManStr,Exponent);
encode_real(_C, {_,Base,_}) ->
exit({error,{asn1, {encode_real_non_supported_encoding, Base}}});
%% base 10
encode_real(C, Real) when is_list(Real) ->
%% The Real string may come in as a NR1, NR2 or NR3 string.
{Mantissa, Exponent} =
case string:tokens(Real,"Ee") of
[NR2] ->
{NR2,0};
[NR3MB,NR3E] ->
%% remove beginning zeros
{NR3MB,list_to_integer(NR3E)}
end,
%% .Decimal | Number | Number.Decimal
ZeroDecimal =
fun("0") -> "";
(L) -> L
end,
{NewMantissa,LenDecimal} =
case Mantissa of
[$.|Dec] ->
NewMan = remove_trailing_zeros(Dec),
{NewMan,length(ZeroDecimal(NewMan))};
_ ->
case string:tokens(Mantissa,",.") of
[Num] -> %% No decimal-mark
{integer_to_list(list_to_integer(Num)),0};
[Num,Dec] ->
NewDec = ZeroDecimal(remove_trailing_zeros(Dec)),
NewMan = integer_to_list(list_to_integer(Num)) ++ NewDec,
{integer_to_list(list_to_integer(NewMan)),
length(NewDec)}
end
end,
encode_real_as_string(C, NewMantissa, Exponent - LenDecimal).
encode_real_as_string(_C, Mantissa, Exponent)
when is_list(Mantissa), is_integer(Exponent) ->
%% Remove trailing zeros in Mantissa and add this to Exponent
TruncMant = remove_trailing_zeros(Mantissa),
ExpIncr = length(Mantissa) - length(TruncMant),
ExpStr = integer_to_list(Exponent + ExpIncr),
ExpBin =
case ExpStr of
"0" ->
<<"E+0">>;
_ ->
ExpB = list_to_binary(ExpStr),
<<$E,ExpB/binary>>
end,
ManBin = list_to_binary(TruncMant),
NR3 = 3,
{<<NR3,ManBin/binary,$.,ExpBin/binary>>,
2 + byte_size(ManBin) + byte_size(ExpBin)}.
remove_trailing_zeros(IntStr) ->
case lists:dropwhile(fun($0)-> true;
(_) -> false
end, lists:reverse(IntStr)) of
[] ->
"0";
ReversedIntStr ->
lists:reverse(ReversedIntStr)
end.
truncate_zeros(Num) ->
truncate_zeros(Num, 0).
truncate_zeros(0, Sum) ->
{0,Sum};
truncate_zeros(M, Sum) ->
case M band 16#f =:= M band 16#e of
true -> truncate_zeros(M bsr 1, Sum+1);
_ -> {M,Sum}
end.
%%============================================================================
%% decode real value
%%
%% decode_real([OctetBufferList], tuple|value, tag|notag) ->
%% {{Mantissa, Base, Exp} | realval | PLUS-INFINITY | MINUS-INFINITY | 0,
%% RestBuff}
%%
%% only for base 2 decoding sofar!!
%%============================================================================
decode_real(Buffer) ->
Sz = byte_size(Buffer),
{RealVal,<<>>,Sz} = decode_real2(Buffer, [], Sz, 0),
RealVal.
decode_real2(Buffer, _C, 0, _RemBytes) ->
{0,Buffer};
decode_real2(Buffer0, _C, Len, RemBytes1) ->
<<First, Buffer2/binary>> = Buffer0,
if
First =:= 2#01000000 -> {'PLUS-INFINITY', Buffer2};
First =:= 2#01000001 -> {'MINUS-INFINITY', Buffer2};
First =:= 1 orelse First =:= 2 orelse First =:= 3 ->
%% charcter string encoding of base 10
{NRx,Rest} = split_binary(Buffer2,Len-1),
{binary_to_list(NRx),Rest,Len};
true ->
%% have some check here to verify only supported bases (2)
%% not base 8 or 16
<<_B7:1,Sign:1,BB:2,_FF:2,EE:2>> = <<First>>,
Base =
case BB of
0 -> 2; % base 2, only one so far
_ -> exit({error,{asn1, {non_supported_base, BB}}})
end,
{FirstLen, {Exp, Buffer3,_Rb2}, RemBytes2} =
case EE of
0 -> {2, decode_integer2(1, Buffer2, RemBytes1), RemBytes1+1};
1 -> {3, decode_integer2(2, Buffer2, RemBytes1), RemBytes1+2};
2 -> {4, decode_integer2(3, Buffer2, RemBytes1), RemBytes1+3};
3 ->
<<ExpLen1,RestBuffer/binary>> = Buffer2,
{ ExpLen1 + 2,
decode_integer2(ExpLen1, RestBuffer, RemBytes1),
RemBytes1+ExpLen1}
end,
%% io:format("FirstLen: ~w, Exp: ~w, Buffer3: ~w ~n",
Length = Len - FirstLen,
<<LongInt:Length/unit:8,RestBuff/binary>> = Buffer3,
{{Mantissa, Buffer4}, RemBytes3} =
if Sign =:= 0 ->
%% io:format("sign plus~n"),
{{LongInt, RestBuff}, 1 + Length};
true ->
%% io:format("sign minus~n"),
{{-LongInt, RestBuff}, 1 + Length}
end,
{{Mantissa, Base, Exp}, Buffer4, RemBytes2+RemBytes3}
end.
encode_pos_integer(0, [B|_Acc]=L) when B < 128 ->
L;
encode_pos_integer(N, Acc) ->
encode_pos_integer(N bsr 8, [N band 16#ff| Acc]).
encode_neg_integer(-1, [B1|_T]=L) when B1 > 127 ->
L;
encode_neg_integer(N, Acc) ->
encode_neg_integer(N bsr 8, [N band 16#ff|Acc]).
%% Val must be >= 0
real_mininum_octets(Val) ->
real_mininum_octets(Val, []).
real_mininum_octets(0, Acc) ->
Acc;
real_mininum_octets(Val, Acc) ->
real_mininum_octets(Val bsr 8, [Val band 16#FF | Acc]).
%% decoding postitive integer values.
decode_integer2(Len, <<0:1,_:7,_Bs/binary>> = Bin, RemovedBytes) ->
<<Int:Len/unit:8,Buffer2/binary>> = Bin,
{Int,Buffer2,RemovedBytes};
%% decoding negative integer values.
decode_integer2(Len, <<1:1,B2:7,Bs/binary>>, RemovedBytes) ->
<<N:Len/unit:8,Buffer2/binary>> = <<B2,Bs/binary>>,
Int = N - (1 bsl (8 * Len - 1)),
{Int,Buffer2,RemovedBytes}.