%% MySQL/OTP – MySQL client library for Erlang/OTP
%% Copyright (C) 2014 Viktor Söderqvist
%%
%% This file is part of MySQL/OTP.
%%
%% MySQL/OTP is free software: you can redistribute it and/or modify it under
%% the terms of the GNU Lesser General Public License as published by the Free
%% Software Foundation, either version 3 of the License, or (at your option)
%% any later version.
%%
%% This program is distributed in the hope that it will be useful, but WITHOUT
%% ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
%% FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
%% more details.
%%
%% You should have received a copy of the GNU Lesser General Public License
%% along with this program. If not, see .
%% @doc This module implements parts of the MySQL client/server protocol.
%%
%% The protocol is described in the document "MySQL Internals" which can be
%% found under "MySQL Documentation: Expert Guides" on http://dev.mysql.com/.
%%
%% TCP communication is not handled in this module. Most of the public functions
%% take funs for data communitaction as parameters.
-module(mysql_protocol).
-export([handshake/5,
query/3,
prepare/3, unprepare/3, execute/4]).
-export_type([sendfun/0, recvfun/0]).
-type sendfun() :: fun((binary()) -> ok).
-type recvfun() :: fun((integer()) -> {ok, binary()}).
%% How much data do we want to send at most?
-define(MAX_BYTES_PER_PACKET, 50000000).
-include("records.hrl").
-include("protocol.hrl").
%% Macros for pattern matching on packets.
-define(ok_pattern, <>).
-define(error_pattern, <>).
-define(eof_pattern, <>).
%% @doc Performs a handshake using the supplied functions for communication.
%% Returns an ok or an error record. Raises errors when various unimplemented
%% features are requested.
%%
%% TODO: Implement setting the database in the handshake. Currently an error
%% occurs if Database is anything other than undefined.
-spec handshake(iodata(), iodata(), iodata() | undefined, sendfun(),
recvfun()) -> #ok{} | #error{}.
handshake(Username, Password, Database, SendFun, RecvFun) ->
SeqNum0 = 0,
Database == undefined orelse error(database_in_handshake),
{ok, HandshakePacket, SeqNum1} = recv_packet(RecvFun, SeqNum0),
Handshake = parse_handshake(HandshakePacket),
Response = build_handshake_response(Handshake, Username, Password),
{ok, SeqNum2} = send_packet(SendFun, Response, SeqNum1),
{ok, ConfirmPacket, _SeqNum3} = recv_packet(RecvFun, SeqNum2),
parse_handshake_confirm(ConfirmPacket).
-spec query(Query :: iodata(), sendfun(), recvfun()) ->
#ok{} | #error{} | #resultset{}.
query(Query, SendFun, RecvFun) ->
Req = <>,
SeqNum0 = 0,
{ok, SeqNum1} = send_packet(SendFun, Req, SeqNum0),
{ok, Resp, SeqNum2} = recv_packet(RecvFun, SeqNum1),
case Resp of
?ok_pattern ->
parse_ok_packet(Resp);
?error_pattern ->
parse_error_packet(Resp);
_ResultSet ->
%% The first packet in a resultset is only the column count.
{ColumnCount, <<>>} = lenenc_int(Resp),
case fetch_resultset(RecvFun, ColumnCount, SeqNum2) of
#error{} = E ->
E;
#resultset{column_definitions = ColDefs, rows = Rows} = R ->
%% Parse the rows according to the 'text protocol'
%% representation.
ColumnTypes = [ColDef#column_definition.type
|| ColDef <- ColDefs],
Rows1 = [decode_text_row(ColumnCount, ColumnTypes, Row)
|| Row <- Rows],
R#resultset{rows = Rows1}
end
end.
%% @doc Prepares a statement.
-spec prepare(iodata(), sendfun(), recvfun()) -> #error{} | #prepared{}.
prepare(Query, SendFun, RecvFun) ->
Req = <>,
{ok, SeqNum1} = send_packet(SendFun, Req, 0),
{ok, Resp, SeqNum2} = recv_packet(RecvFun, SeqNum1),
case Resp of
?error_pattern ->
parse_error_packet(Resp);
<> ->
%% This was the first packet.
%% Now: Parameter Definition Block. The parameter definitions don't
%% contain any useful data at all. They are always TYPE_VAR_STRING
%% with charset 'binary' so we have to select a type ourselves for
%% the parameters we have in execute/4.
{_ParamDefs, SeqNum3} =
fetch_column_definitions_if_any(NumParams, RecvFun, SeqNum2),
%% Column Definition Block. We get column definitions in execute
%% too, so we don't need them here. We *could* store them to be able
%% to provide the user with some info about a prepared statement.
{_ColDefs, _SeqNum4} =
fetch_column_definitions_if_any(NumColumns, RecvFun, SeqNum3),
#prepared{statement_id = StmtId,
param_count = NumParams,
warning_count = WarningCount}
end.
%% @doc Deallocates a prepared statement.
-spec unprepare(#prepared{}, sendfun(), recvfun()) -> ok.
unprepare(#prepared{statement_id = Id}, SendFun, _RecvFun) ->
{ok, _SeqNum} = send_packet(SendFun, <>, 0),
ok.
%% @doc Executes a prepared statement.
-spec execute(#prepared{}, [term()], sendfun(), recvfun()) -> #resultset{}.
execute(#prepared{statement_id = Id, param_count = ParamCount}, ParamValues,
SendFun, RecvFun) when ParamCount == length(ParamValues) ->
%% Flags Constant Name
%% 0x00 CURSOR_TYPE_NO_CURSOR
%% 0x01 CURSOR_TYPE_READ_ONLY
%% 0x02 CURSOR_TYPE_FOR_UPDATE
%% 0x04 CURSOR_TYPE_SCROLLABLE
Flags = 0,
Req0 = <>,
Req = case ParamCount of
0 ->
Req0;
_ ->
%% We can't use the parameter types returned by the prepare call.
%% They are all reported as ?TYPE_VAR_STRING with character
%% set 'binary'.
NullBitMap = build_null_bitmap(ParamValues),
%% What does it mean to *not* bind new params? To use the same
%% params as last time? Right now we always bind params each time.
NewParamsBoundFlag = 1,
Req1 = <>,
%% For each value, first append type and signedness (16#80 signed or
%% 00 unsigned) for all values and then the binary encoded values.
EncodedParams = lists:map(fun encode_param/1, ParamValues),
{TypesAndSigns, EncValues} = lists:unzip(EncodedParams),
iolist_to_binary([Req1, TypesAndSigns, EncValues])
end,
{ok, SeqNum1} = send_packet(SendFun, Req, 0),
{ok, Resp, SeqNum2} = recv_packet(RecvFun, SeqNum1),
case Resp of
?ok_pattern ->
parse_ok_packet(Resp);
?error_pattern ->
parse_error_packet(Resp);
_ResultPacket ->
%% The first packet in a resultset is only the column count.
{ColumnCount, <<>>} = lenenc_int(Resp),
case fetch_resultset(RecvFun, ColumnCount, SeqNum2) of
#error{} = E ->
%% TODO: Find a way to get here and write a testcase.
%% This can happen for the text protocol but maybe not for
%% the binary protocol.
E;
#resultset{column_definitions = ColDefs, rows = Rows} = R ->
%% Parse the rows according to the 'binary protocol'
%% representation.
ColumnTypes = [ColDef#column_definition.type
|| ColDef <- ColDefs],
Rows1 = [decode_binary_row(ColumnCount, ColumnTypes, Row)
|| Row <- Rows],
R#resultset{rows = Rows1}
end
end.
%% --- internal ---
%% @doc Parses a handshake. This is the first thing that comes from the server
%% when connecting. If an unsupported version or variant of the protocol is used
%% an error is raised.
-spec parse_handshake(binary()) -> #handshake{}.
parse_handshake(<<10, Rest/binary>>) ->
%% Protocol version 10.
{ServerVersion, Rest1} = nulterm_str(Rest),
<> = Rest1,
Capabilities = CapabilitiesLower + 16#10000 * CapabilitiesUpper,
Len = case AuthPluginDataLength of
0 -> 13; %% Server has not CLIENT_PLUGIN_AUTH
K -> K - 8 %% Part 2 length = Total length minus the 8 bytes in part 1.
end,
<> = Rest3,
AuthPluginData = <>,
%% "Due to Bug#59453 the auth-plugin-name is missing the terminating
%% NUL-char in versions prior to 5.5.10 and 5.6.2."
%% Strip the final NUL byte if any.
NameLen = size(AuthPluginName) - 1,
AuthPluginName1 = case AuthPluginName of
<> -> NameNoNul;
_ -> AuthPluginName
end,
#handshake{server_version = ServerVersion,
connection_id = ConnectionId,
capabilities = Capabilities,
character_set = CharacterSet,
status = StatusFlags,
auth_plugin_data = AuthPluginData,
auth_plugin_name = AuthPluginName1};
parse_handshake(<>) when Protocol /= 10 ->
error(unknown_protocol).
%% @doc The response sent by the client to the server after receiving the
%% initial handshake from the server
-spec build_handshake_response(#handshake{}, iodata(), iodata()) -> binary().
build_handshake_response(Handshake, Username, Password) ->
%% We require these capabilities. Make sure the server handles them.
CapabilityFlags = ?CLIENT_PROTOCOL_41 bor
?CLIENT_TRANSACTIONS bor
?CLIENT_SECURE_CONNECTION,
Handshake#handshake.capabilities band CapabilityFlags == CapabilityFlags
orelse error(old_server_version),
Hash = case Handshake#handshake.auth_plugin_name of
<<>> ->
%% Server doesn't know auth plugins
hash_password(Password, Handshake#handshake.auth_plugin_data);
<<"mysql_native_password">> ->
hash_password(Password, Handshake#handshake.auth_plugin_data);
UnknownAuthMethod ->
error({auth_method, UnknownAuthMethod})
end,
HashLength = size(Hash),
CharacterSet = ?UTF8,
UsernameUtf8 = unicode:characters_to_binary(Username),
<>.
%% @doc Handles the second packet from the server, when we have replied to the
%% initial handshake. Returns an error if the server returns an error. Raises
%% an error if unimplemented features are required.
-spec parse_handshake_confirm(binary()) -> #ok{} | #error{}.
parse_handshake_confirm(Packet) ->
case Packet of
?ok_pattern ->
%% Connection complete.
parse_ok_packet(Packet);
?error_pattern ->
%% "Insufficient Client Capabilities"
parse_error_packet(Packet);
<> ->
%% "Old Authentication Method Switch Request Packet consisting of a
%% single 0xfe byte. It is sent by server to request client to
%% switch to Old Password Authentication if CLIENT_PLUGIN_AUTH
%% capability is not supported (by either the client or the server)"
error(old_auth);
<> ->
%% "Authentication Method Switch Request Packet. If both server and
%% client support CLIENT_PLUGIN_AUTH capability, server can send
%% this packet to ask client to use another authentication method."
error(auth_method_switch)
end.
%% Fetches packets until a
-spec fetch_resultset(recvfun(), integer(), integer()) ->
#resultset{} | #error{}.
fetch_resultset(RecvFun, FieldCount, SeqNum) ->
{ok, ColDefs, SeqNum1} = fetch_column_definitions(RecvFun, SeqNum,
FieldCount, []),
{ok, DelimiterPacket, SeqNum2} = recv_packet(RecvFun, SeqNum1),
#eof{} = parse_eof_packet(DelimiterPacket),
case fetch_resultset_rows(RecvFun, SeqNum2, []) of
{ok, Rows, _SeqNum3} ->
#resultset{column_definitions = ColDefs, rows = Rows};
#error{} = E ->
E
end.
%% @doc Receives NumLeft packets and parses them as column definitions.
%% TODO: Don't parse them here. That's a sepatate thing we not always need to
%% do.
-spec fetch_column_definitions(recvfun(), SeqNum :: integer(),
NumLeft :: integer(), Acc :: [tuple()]) ->
{ok, [tuple()], NextSeqNum :: integer()}.
fetch_column_definitions(RecvFun, SeqNum, NumLeft, Acc) when NumLeft > 0 ->
{ok, Packet, SeqNum1} = recv_packet(RecvFun, SeqNum),
ColDef = parse_column_definition(Packet),
fetch_column_definitions(RecvFun, SeqNum1, NumLeft - 1, [ColDef | Acc]);
fetch_column_definitions(_RecvFun, SeqNum, 0, Acc) ->
{ok, lists:reverse(Acc), SeqNum}.
%% @doc Fetches rows in a result set. There is a packet per row. The row packets
%% are not decoded. This function can be used for both the binary and the text
%% protocol result sets.
-spec fetch_resultset_rows(recvfun(), SeqNum :: integer(), Acc) ->
{ok, Rows, integer()} | #error{}
when Acc :: [binary()],
Rows :: [binary()].
fetch_resultset_rows(RecvFun, SeqNum, Acc) ->
{ok, Packet, SeqNum1} = recv_packet(RecvFun, SeqNum),
case Packet of
?error_pattern ->
parse_error_packet(Packet);
?eof_pattern ->
{ok, lists:reverse(Acc), SeqNum1};
Row ->
fetch_resultset_rows(RecvFun, SeqNum1, [Row | Acc])
end.
%% -- both text and binary protocol --
%% Parses a packet containing a column definition (part of a result set)
parse_column_definition(Data) ->
{<<"def">>, Rest1} = lenenc_str(Data), %% catalog (always "def")
{_Schema, Rest2} = lenenc_str(Rest1), %% schema-name
{_Table, Rest3} = lenenc_str(Rest2), %% virtual table-name
{_OrgTable, Rest4} = lenenc_str(Rest3), %% physical table-name
{Name, Rest5} = lenenc_str(Rest4), %% virtual column name
{_OrgName, Rest6} = lenenc_str(Rest5), %% physical column name
{16#0c, Rest7} = lenenc_int(Rest6), %% length of the following fields
%% (always 0x0c)
<> = Rest7, %% - 0x00 to 0x51 for decimals
%% Here, if command was COM_FIELD_LIST {
%% default values: lenenc_str
%% }
<<>> = Rest8,
#column_definition{name = Name, type = ColumnType, charset = Charset}.
%% -- text protocol --
-spec decode_text_row(NumColumns :: integer(), ColumnTypes :: integer(),
Data :: binary()) -> [term()].
decode_text_row(_NumColumns, ColumnTypes, Data) ->
decode_text_row_acc(ColumnTypes, Data, []).
%% parses Data using ColDefs and builds the values Acc.
decode_text_row_acc([Type | Types], Data, Acc) ->
case Data of
<<16#fb, Rest/binary>> ->
%% NULL
decode_text_row_acc(Types, Rest, [null | Acc]);
_ ->
%% Every thing except NULL
{Text, Rest} = lenenc_str(Data),
Term = decode_text(Type, Text),
decode_text_row_acc(Types, Rest, [Term | Acc])
end;
decode_text_row_acc([], <<>>, Acc) ->
lists:reverse(Acc).
%% @doc When receiving data in the text protocol, we get everything as binaries
%% (except NULL). This function is used to parse these strings values.
decode_text(_, null) ->
%% NULL is the only value not represented as a binary.
null;
decode_text(T, Text)
when T == ?TYPE_TINY; T == ?TYPE_SHORT; T == ?TYPE_LONG; T == ?TYPE_LONGLONG;
T == ?TYPE_INT24; T == ?TYPE_YEAR; T == ?TYPE_BIT ->
%% For BIT, do we want bitstring, int or binary?
binary_to_integer(Text);
decode_text(T, Text)
when T == ?TYPE_DECIMAL; T == ?TYPE_NEWDECIMAL; T == ?TYPE_VARCHAR;
T == ?TYPE_ENUM; T == ?TYPE_TINY_BLOB; T == ?TYPE_MEDIUM_BLOB;
T == ?TYPE_LONG_BLOB; T == ?TYPE_BLOB; T == ?TYPE_VAR_STRING;
T == ?TYPE_STRING; T == ?TYPE_GEOMETRY ->
Text;
decode_text(?TYPE_DATE, <>) ->
{binary_to_integer(Y), binary_to_integer(M), binary_to_integer(D)};
decode_text(?TYPE_TIME, Text) ->
{match, [Sign, Hbin, Mbin, Sbin, Frac]} =
re:run(Text,
<<"^(-?)(\\d+):(\\d+):(\\d+)(\\.?\\d*)$">>,
[{capture, all_but_first, binary}]),
H = binary_to_integer(Hbin),
M = binary_to_integer(Mbin),
S = binary_to_integer(Sbin),
IsNeg = Sign == <<"-">>,
Fraction = case Frac of
<<>> -> 0;
_ when not IsNeg -> binary_to_float(<<"0", Frac/binary>>);
_ when IsNeg -> 1 - binary_to_float(<<"0", Frac/binary>>)
end,
Sec1 = H * 3600 + M * 60 + S,
Sec2 = if IsNeg -> -Sec1; true -> Sec1 end,
Sec3 = if IsNeg and (Fraction /= 0) -> Sec2 - 1;
true -> Sec2
end,
{Days, {Hours, Minutes, Seconds}} = calendar:seconds_to_daystime(Sec3),
{Days, {Hours, Minutes, Seconds + Fraction}};
decode_text(T, <>)
when T == ?TYPE_TIMESTAMP; T == ?TYPE_DATETIME ->
%% Without fractions.
{{binary_to_integer(Y), binary_to_integer(M), binary_to_integer(D)},
{binary_to_integer(H), binary_to_integer(Mi), binary_to_integer(S)}};
decode_text(T, <>)
when T == ?TYPE_TIMESTAMP; T == ?TYPE_DATETIME ->
%% With fractions.
{{binary_to_integer(Y), binary_to_integer(M), binary_to_integer(D)},
{binary_to_integer(H), binary_to_integer(Mi), binary_to_float(FloatS)}};
decode_text(T, Text) when T == ?TYPE_FLOAT; T == ?TYPE_DOUBLE ->
try binary_to_float(Text)
catch error:badarg ->
try binary_to_integer(Text) of
Int -> float(Int)
catch error:badarg ->
%% It is something like "4e75" that must be turned into "4.0e75"
binary_to_float(binary:replace(Text, <<"e">>, <<".0e">>))
end
end;
decode_text(?TYPE_SET, <<>>) ->
sets:new();
decode_text(?TYPE_SET, Text) ->
sets:from_list(binary:split(Text, <<",">>, [global])).
%% -- binary protocol --
%% @doc If NumColumns is non-zero, fetches this number of column definitions
%% and an EOF packet. Used by prepare/3.
fetch_column_definitions_if_any(0, _RecvFun, SeqNum) ->
{[], SeqNum};
fetch_column_definitions_if_any(N, RecvFun, SeqNum) ->
{ok, Defs, SeqNum1} = fetch_column_definitions(RecvFun, SeqNum, N, []),
{ok, ?eof_pattern, SeqNum2} = recv_packet(RecvFun, SeqNum1),
{Defs, SeqNum2}.
%% @doc Decodes a packet representing a row in a binary result set.
%% It consists of a 0 byte, then a null bitmap, then the values.
%% Returns a list of length NumColumns with terms of appropriate types for each
%% MySQL type in ColumnTypes.
-spec decode_binary_row(NumColumns :: integer(), ColumnTypes :: [integer()],
Data :: binary()) -> [term()].
decode_binary_row(NumColumns, ColumnTypes, <<0, Data/binary>>) ->
{NullBitMap, Rest} = null_bitmap_decode(NumColumns, Data, 2),
decode_binary_row_acc(ColumnTypes, NullBitMap, Rest, []).
%% @doc Accumulating helper for decode_binary_row/3.
decode_binary_row_acc([_ | Types], <<1:1, NullBitMap/bitstring>>, Data, Acc) ->
%% NULL
decode_binary_row_acc(Types, NullBitMap, Data, [null | Acc]);
decode_binary_row_acc([Type | Types], <<0:1, NullBitMap/bitstring>>, Data,
Acc) ->
%% Not NULL
{Term, Rest} = decode_binary(Type, Data),
decode_binary_row_acc(Types, NullBitMap, Rest, [Term | Acc]);
decode_binary_row_acc([], _, <<>>, Acc) ->
lists:reverse(Acc).
%% @doc Decodes a null bitmap as stored by MySQL and returns it in a strait
%% bitstring counting bits from left to right in a tuple with remaining data.
%%
%% In the MySQL null bitmap the bits are stored counting bytes from the left and
%% bits within each byte from the right. (Sort of little endian.)
-spec null_bitmap_decode(NumColumns :: integer(), BitOffset :: integer(),
Data :: binary()) ->
{NullBitstring :: bitstring(), Rest :: binary()}.
null_bitmap_decode(NumColumns, Data, BitOffset) ->
%% Binary shift right by 3 is equivallent to integer division by 8.
BitMapLength = (NumColumns + BitOffset + 7) bsr 3,
<> = Data,
<<_:BitOffset, NullBitstring:NumColumns/bitstring, _/bitstring>> =
<< <<(reverse_byte(B))/binary>> || <> <= NullBitstring0 >>,
{NullBitstring, Rest}.
%% @doc The reverse of null_bitmap_decode/3. The number of columns is taken to
%% be the number of bits in NullBitstring. Returns the MySQL null bitmap as a
%% binary (i.e. full bytes). BitOffset is the number of unused bits that should
%% be inserted before the other bits.
-spec null_bitmap_encode(bitstring(), integer()) -> binary().
null_bitmap_encode(NullBitstring, BitOffset) ->
PayloadLength = bit_size(NullBitstring) + BitOffset,
%% Round up to a multiple of 8.
BitMapLength = (PayloadLength + 7) band bnot 7,
PadBitsLength = BitMapLength - PayloadLength,
PaddedBitstring = <<0:BitOffset, NullBitstring/bitstring, 0:PadBitsLength>>,
<< <<(reverse_byte(B))/binary>> || <> <= PaddedBitstring >>.
%% Reverses the bits in a byte.
reverse_byte(<>) ->
<>.
%% @doc Used for executing prepared statements. The bit offset whould be 0 in
%% this case.
-spec build_null_bitmap([any()]) -> binary().
build_null_bitmap(Values) ->
Bits = << <<(case V of null -> 1; _ -> 0 end):1>> || V <- Values >>,
null_bitmap_encode(Bits, 0).
%% Decodes a value as received in the 'binary protocol' result set.
%%
%% The types are type constants for the binary protocol, such as
%% ProtocolBinary::MYSQL_TYPE_STRING. In the guide "MySQL Internals" these are
%% not listed, but we assume that are the same as for the text protocol.
-spec decode_binary(Type :: integer(), Data :: binary()) ->
{Term :: term(), Rest :: binary()}.
decode_binary(T, Data)
when T == ?TYPE_STRING; T == ?TYPE_VARCHAR; T == ?TYPE_VAR_STRING;
T == ?TYPE_ENUM; T == ?TYPE_SET; T == ?TYPE_LONG_BLOB;
T == ?TYPE_MEDIUM_BLOB; T == ?TYPE_BLOB; T == ?TYPE_TINY_BLOB;
T == ?TYPE_GEOMETRY; T == ?TYPE_BIT; T == ?TYPE_DECIMAL;
T == ?TYPE_NEWDECIMAL ->
lenenc_str(Data);
decode_binary(?TYPE_LONGLONG, <>) ->
{Value, Rest};
decode_binary(T, <>)
when T == ?TYPE_LONG; T == ?TYPE_INT24 ->
{Value, Rest};
decode_binary(T, <>)
when T == ?TYPE_SHORT; T == ?TYPE_YEAR ->
{Value, Rest};
decode_binary(?TYPE_TINY, <>) ->
{Value, Rest};
decode_binary(?TYPE_DOUBLE, <>) ->
{Value, Rest};
decode_binary(?TYPE_FLOAT, <>) ->
%% There is a precision loss when storing and fetching a 32-bit float.
%% In the text protocol, it is obviously rounded. Storing 3.14 in a FLOAT
%% column and fetching it using the text protocol, we get "3.14" which we
%% parse to the Erlang double as close as possible to 3.14. Fetching the
%% same value as a binary 32-bit float, we get 3.140000104904175. To achieve
%% the same rounding after receiving it as a 32-bit float, we try to do the
%% same rounding here as MySQL does when sending it over the text protocol.
%%
%% This comment explains the idea:
%%
%% Posted by Geoffrey Downs on March 10 2011 10:26am
%%
%% Following up... I *think* this is correct for the default float
%% columns in mysql:
%%
%% var yourNumber = some floating point value
%% max decimal precision = 10 ^ (-5 + floor(yourNumber log 10))
%% So:
%% 0 < x < 10 -> max precision is 0.00001
%% 10 <= x < 100 -> max precision is 0.0001
%% 100 <= x < 1000 -> max precision is 0.001
%% etc.
%%
%% (From http://dev.mysql.com/doc/refman/5.7/en/problems-with-float.html
%% fetched 10 Nov 2014)
%%
%% The above is almost correct, except for the example in the interval
%% 0 < x < 1. There are 6 significant digits also for these numbers.
%%
%% Now, instead of P = 0.00001 we want the inverse 100000.0 but if we
%% compute Factor = 1 / P we get a precision loss, so instead we do this:
Factor = math:pow(10, floor(6 - math:log10(abs(Value)))),
RoundedValue = round(Value * Factor) / Factor,
{RoundedValue, Rest};
decode_binary(?TYPE_DATE, <>) ->
%% Coded in the same way as DATETIME and TIMESTAMP below, but returned in
%% a simple triple.
case {Length, Data} of
{0, _} -> {{0, 0, 0}, Data};
{4, <>} -> {{Y, M, D}, Rest}
end;
decode_binary(T, <>)
when T == ?TYPE_DATETIME; T == ?TYPE_TIMESTAMP ->
%% length (1) -- number of bytes following (valid values: 0, 4, 7, 11)
case {Length, Data} of
{0, _} ->
{{{0, 0, 0}, {0, 0, 0}}, Data};
{4, <>} ->
{{{Y, M, D}, {0, 0, 0}}, Rest};
{7, <>} ->
{{{Y, M, D}, {H, Mi, S}}, Rest};
{11, <>} ->
{{{Y, M, D}, {H, Mi, S + 0.000001 * Micro}}, Rest}
end;
decode_binary(?TYPE_TIME, <>) ->
%% length (1) -- number of bytes following (valid values: 0, 8, 12)
%% is_negative (1) -- (1 if minus, 0 for plus)
%% days (4) -- days
%% hours (1) -- hours
%% minutes (1) -- minutes
%% seconds (1) -- seconds
%% micro_seconds (4) -- micro-seconds
case {Length, Data} of
{0, _} ->
{{0, {0, 0, 0}}, Data};
{8, <<0, D:32/little, H, M, S, Rest/binary>>} ->
{{D, {H, M, S}}, Rest};
{12, <<0, D:32/little, H, M, S, Micro:32/little, Rest/binary>>} ->
{{D, {H, M, S + 0.000001 * Micro}}, Rest};
{8, <<1, D:32/little, H, M, S, Rest/binary>>} ->
%% Negative time. Example: '-00:00:01' --> {-1,{23,59,59}}
Seconds = ((D * 24 + H) * 60 + M) * 60 + S,
%Seconds = D * 86400 + calendar:time_to_seconds({H, M, S}),
{calendar:seconds_to_daystime(-Seconds), Rest};
{12, <<1, D:32/little, H, M, S, Micro:32/little, Rest/binary>>}
when Micro > 0 ->
%% Negate and convert to seconds, excl fractions
Seconds = -(((D * 24 + H) * 60 + M) * 60 + S),
%Seconds = -D * 86400 - calendar:time_to_seconds({H, M, S}),
%% Subtract 1 second for the fractions
{Days, {Hours, Minutes, Sec}} =
calendar:seconds_to_daystime(Seconds - 1),
%% Adding the fractions to Sec again makes it a float
{{Days, {Hours, Minutes, Sec + 1 - 0.000001 * Micro}}, Rest}
end.
%% @doc Like trunc/1 but towards negative infinity instead of towards zero.
floor(Value) ->
Trunc = trunc(Value),
if
Trunc =< Value -> Trunc;
Trunc > Value -> Trunc - 1 %% for negative values
end.
%% @doc Encodes a term reprenting av value as a binary for use in the binary
%% protocol. As this is used to encode parameters for prepared statements, the
%% encoding is in its required form, namely `<>'.
%%
%% TODO: Maybe change Erlang representation of BIT to `<<_:1>>'.
-spec encode_param(term()) -> {TypeAndSign :: binary(), Data :: binary()}.
encode_param(null) ->
{<>, <<>>};
encode_param(Value) when is_binary(Value) ->
EncLength = lenenc_int_encode(byte_size(Value)),
{<>, <>};
encode_param(Value) when is_integer(Value), Value >= 0 ->
%% We send positive integers with the 'unsigned' flag set.
if
Value =< 16#ff ->
{<>, <>};
Value =< 16#ffff ->
{<>, <>};
Value =< 16#ffffffff ->
{<>, <>};
Value =< 16#ffffffffffffffff ->
{<>, <>};
true ->
%% If larger than a 64-bit int we send it as a string. MySQL does
%% silently cast strings in aithmetic expressions. Also, DECIMALs
%% are always sent as strings.
encode_param(integer_to_binary(Value))
end;
encode_param(Value) when is_integer(Value), Value < 0 ->
if
Value >= -16#80 ->
{<>, <>};
Value >= -16#8000 ->
{<>, <>};
Value >= -16#80000000 ->
{<>, <>};
Value >= -16#8000000000000000 ->
{<>, <>};
true ->
encode_param(integer_to_binary(Value))
end;
encode_param(Value) when is_float(Value) ->
{<>, <>};
encode_param({Y, M, D}) ->
%% calendar:date()
{<>, <<4, Y:16/little, M, D>>};
encode_param({{Y, M, D}, {0, 0, 0}}) ->
%% Datetime at midnight
{<>, <<4, Y:16/little, M, D>>};
encode_param({{Y, M, D}, {H, Mi, S}}) when is_integer(S) ->
%% calendar:datetime()
{<>, <<7, Y:16/little, M, D, H, Mi, S>>};
encode_param({{Y, M, D}, {H, Mi, S}}) when is_float(S) ->
%% calendar:datetime() with a float for seconds. This way it looks very
%% similar to a datetime. Microseconds in MySQL timestamps are possible but
%% not very common.
Sec = trunc(S),
Micro = round(1000000 * (S - Sec)),
{<>, <<11, Y:16/little, M, D, H, Mi, Sec,
Micro:32/little>>};
encode_param({D, {H, M, S}}) when is_integer(S), D >= 0 ->
%% calendar:seconds_to_daystime()
{<>, <<8, 0, D:32/little, H, M, S>>};
encode_param({D, {H, M, S}}) when is_integer(S), D < 0 ->
%% Convert to seconds, negate and convert back to daystime form.
%% Then set the minus flag.
Seconds = ((D * 24 + H) * 60 + M) * 60 + S,
{D1, {H1, M1, S1}} = calendar:seconds_to_daystime(-Seconds),
{<>, <<8, 1, D1:32/little, H1, M1, S1>>};
encode_param({D, {H, M, S}}) when is_float(S), D >= 0 ->
S1 = trunc(S),
Micro = round(1000000 * (S - S1)),
{<>, <<12, 0, D:32/little, H, M, S1, Micro:32/little>>};
encode_param({D, {H, M, S}}) when is_float(S), S > 0.0, D < 0 ->
IntS = trunc(S),
Micro = round(1000000 * (1 - S + IntS)),
Seconds = (D * 24 + H) * 3600 + M * 60 + IntS + 1,
{D1, {M1, H1, S1}} = calendar:seconds_to_daystime(-Seconds),
{<>, <<12, 1, D1:32/little, H1, M1, S1, Micro:32/little>>};
encode_param({D, {H, M, 0.0}}) ->
encode_param({D, {H, M, 0}}).
%% -- Protocol basics: packets --
%% @doc Wraps Data in packet headers, sends it by calling SendFun and returns
%% {ok, SeqNum1} where SeqNum1 is the next sequence number.
-spec send_packet(sendfun(), Data :: binary(), SeqNum :: integer()) ->
{ok, NextSeqNum :: integer()}.
send_packet(SendFun, Data, SeqNum) ->
{WithHeaders, SeqNum1} = add_packet_headers(Data, SeqNum),
ok = SendFun(WithHeaders),
{ok, SeqNum1}.
%% @doc Receives data by calling RecvFun and removes the packet headers. Returns
%% the packet contents and the next packet sequence number.
-spec recv_packet(RecvFun :: recvfun(), SeqNum :: integer()) ->
{ok, Data :: binary(), NextSeqNum :: integer()}.
recv_packet(RecvFun, SeqNum) ->
recv_packet(RecvFun, SeqNum, <<>>).
%% @doc Receives data by calling RecvFun and removes packet headers. Returns the
%% data and the next packet sequence number.
-spec recv_packet(RecvFun :: recvfun(), ExpectSeqNum :: integer(),
Acc :: binary()) ->
{ok, Data :: binary(), NextSeqNum :: integer()}.
recv_packet(RecvFun, ExpectSeqNum, Acc) ->
{ok, Header} = RecvFun(4),
{Size, ExpectSeqNum, More} = parse_packet_header(Header),
{ok, Body} = RecvFun(Size),
Acc1 = <>,
NextSeqNum = (ExpectSeqNum + 1) band 16#ff,
case More of
false -> {ok, Acc1, NextSeqNum};
true -> recv_packet(RecvFun, NextSeqNum, Acc1)
end.
%% @doc Parses a packet header (32 bits) and returns a tuple.
%%
%% The client should first read a header and parse it. Then read PacketLength
%% bytes. If there are more packets, read another header and read a new packet
%% length of payload until there are no more packets. The seq num should
%% increment from 0 and may wrap around at 255 back to 0.
%%
%% When all packets are read and the payload of all packets are concatenated, it
%% can be parsed using parse_response/1, etc. depending on what type of response
%% is expected.
-spec parse_packet_header(PackerHeader :: binary()) ->
{PacketLength :: integer(),
SeqNum :: integer(),
MorePacketsExist :: boolean()}.
parse_packet_header(<>) ->
{PacketLength, SeqNum, PacketLength == 16#ffffff}.
%% @doc Splits a packet body into chunks and wraps them in headers. The
%% resulting list is ready to sent to the socket.
-spec add_packet_headers(PacketBody :: iodata(), SeqNum :: integer()) ->
{PacketWithHeaders :: iodata(), NextSeqNum :: integer()}.
add_packet_headers(PacketBody, SeqNum) ->
Bin = iolist_to_binary(PacketBody),
Size = size(Bin),
SeqNum1 = (SeqNum + 1) rem 16#100,
%% Todo: implement the case when Size >= 16#ffffff.
if Size < 16#ffffff ->
{[<>, Bin], SeqNum1}
end.
-spec parse_ok_packet(binary()) -> #ok{}.
parse_ok_packet(<>) ->
{AffectedRows, Rest1} = lenenc_int(Rest),
{InsertId, Rest2} = lenenc_int(Rest1),
<> = Rest2,
%% We have CLIENT_PROTOCOL_41 but not CLIENT_SESSION_TRACK enabled. The
%% protocol is conditional. This is from the protocol documentation:
%%
%% if capabilities & CLIENT_PROTOCOL_41 {
%% int<2> status_flags
%% int<2> warning_count
%% } elseif capabilities & CLIENT_TRANSACTIONS {
%% int<2> status_flags
%% }
%% if capabilities & CLIENT_SESSION_TRACK {
%% string info
%% if status_flags & SERVER_SESSION_STATE_CHANGED {
%% string session_state_changes
%% }
%% } else {
%% string info
%% }
#ok{affected_rows = AffectedRows,
insert_id = InsertId,
status = StatusFlags,
warning_count = WarningCount,
msg = Msg}.
-spec parse_error_packet(binary()) -> #error{}.
parse_error_packet(<>) ->
%% Error, 4.1 protocol.
%% (Older protocol: <>)
#error{code = ErrNo, state = SQLState, msg = Msg}.
-spec parse_eof_packet(binary()) -> #eof{}.
parse_eof_packet(<>) ->
%% EOF packet, 4.1 protocol.
%% (Older protocol: <>)
#eof{status = StatusFlags, warning_count = NumWarnings}.
-spec hash_password(Password :: iodata(), Salt :: binary()) -> Hash :: binary().
hash_password(Password, Salt) ->
%% From the "MySQL Internals" manual:
%% SHA1( password ) XOR SHA1( "20-bytes random data from server"
%% SHA1( SHA1( password ) ) )
%% ----
%% Make sure the salt is exactly 20 bytes.
%%
%% The auth data is obviously nul-terminated. For the "native" auth
%% method, it should be a 20 byte salt, so let's trim it in this case.
Salt1 = case Salt of
<> -> SaltNoNul;
_ when size(Salt) == 20 -> Salt
end,
%% Hash as described above.
<> = Hash1 = crypto:hash(sha, Password),
Hash2 = crypto:hash(sha, Hash1),
<> = crypto:hash(sha, <>),
<<(Hash1Num bxor Hash3Num):160>>.
%% --- Lowlevel: variable length integers and strings ---
%% lenenc_int/1 decodes length-encoded-integer values
-spec lenenc_int(Input :: binary()) -> {Value :: integer(), Rest :: binary()}.
lenenc_int(<>) when Value < 251 -> {Value, Rest};
lenenc_int(<<16#fc:8, Value:16/little, Rest/binary>>) -> {Value, Rest};
lenenc_int(<<16#fd:8, Value:24/little, Rest/binary>>) -> {Value, Rest};
lenenc_int(<<16#fe:8, Value:64/little, Rest/binary>>) -> {Value, Rest}.
%% Length-encoded-integer encode. Appends the encoded value to Acc.
%% Values not representable in 64 bits are not accepted.
-spec lenenc_int_encode(0..16#ffffffffffffffff) -> binary().
lenenc_int_encode(Value) when Value < 0 ->
error(badarg);
lenenc_int_encode(Value) when Value < 251 ->
<>;
lenenc_int_encode(Value) when Value =< 16#ffff ->
<<16#fc, Value:16/little>>;
lenenc_int_encode(Value) when Value =< 16#ffffff ->
<<16#fd, Value:24/little>>;
lenenc_int_encode(Value) when Value =< 16#ffffffffffffffff ->
<<16#fe, Value:64/little>>.
%% lenenc_str/1 decodes length-encoded-string values
-spec lenenc_str(Input :: binary()) -> {String :: binary(), Rest :: binary()}.
lenenc_str(Bin) ->
{Length, Rest} = lenenc_int(Bin),
<> = Rest,
{String, Rest1}.
%% nts/1 decodes a nul-terminated string
-spec nulterm_str(Input :: binary()) -> {String :: binary(), Rest :: binary()}.
nulterm_str(Bin) ->
[String, Rest] = binary:split(Bin, <<0>>),
{String, Rest}.
-ifdef(TEST).
-include_lib("eunit/include/eunit.hrl").
-include("server_status.hrl").
%% Testing some of the internal functions, mostly the cases we don't cover in
%% other tests.
decode_text_test() ->
%% Int types
lists:foreach(fun (T) -> ?assertEqual(1, decode_text(T, <<"1">>)) end,
[?TYPE_TINY, ?TYPE_SHORT, ?TYPE_LONG, ?TYPE_LONGLONG,
?TYPE_INT24, ?TYPE_YEAR, ?TYPE_BIT]),
%% Floating point and decimal numbers
lists:foreach(fun (T) -> ?assertEqual(3.0, decode_text(T, <<"3.0">>)) end,
[?TYPE_FLOAT, ?TYPE_DOUBLE]),
%% Decimal types
lists:foreach(fun (T) ->
?assertEqual(<<"3.0">>, decode_text(T, <<"3.0">>))
end,
[?TYPE_DECIMAL, ?TYPE_NEWDECIMAL]),
?assertEqual(3.0, decode_text(?TYPE_FLOAT, <<"3">>)),
?assertEqual(30.0, decode_text(?TYPE_FLOAT, <<"3e1">>)),
?assertEqual(3, decode_text(?TYPE_LONG, <<"3">>)),
%% Date and time
?assertEqual({2014, 11, 01}, decode_text(?TYPE_DATE, <<"2014-11-01">>)),
?assertEqual({0, {23, 59, 01}}, decode_text(?TYPE_TIME, <<"23:59:01">>)),
?assertEqual({{2014, 11, 01}, {23, 59, 01}},
decode_text(?TYPE_DATETIME, <<"2014-11-01 23:59:01">>)),
?assertEqual({{2014, 11, 01}, {23, 59, 01}},
decode_text(?TYPE_TIMESTAMP, <<"2014-11-01 23:59:01">>)),
%% Strings and blobs
lists:foreach(fun (T) ->
?assertEqual(<<"x">>, decode_text(T, <<"x">>))
end,
[?TYPE_VARCHAR, ?TYPE_ENUM, ?TYPE_TINY_BLOB,
?TYPE_MEDIUM_BLOB, ?TYPE_LONG_BLOB, ?TYPE_BLOB,
?TYPE_VAR_STRING, ?TYPE_STRING, ?TYPE_GEOMETRY]),
%% Set
?assertEqual(sets:from_list([<<"b">>, <<"a">>]),
decode_text(?TYPE_SET, <<"a,b">>)),
?assertEqual(sets:from_list([]), decode_text(?TYPE_SET, <<>>)),
%% NULL
?assertEqual(null, decode_text(?TYPE_FLOAT, null)),
ok.
decode_binary_test() ->
%% Test the special rounding we apply to (single precision) floats.
%?assertEqual({1.0, <<>>},
% decode_binary(?TYPE_FLOAT, <<1.0:32/float-little>>)),
%?assertEqual({0.2, <<>>},
% decode_binary(?TYPE_FLOAT, <<0.2:32/float-little>>)),
%?assertEqual({-33.3333, <<>>},
% decode_binary(?TYPE_FLOAT, <<-33.333333:32/float-little>>)),
%?assertEqual({0.000123457, <<>>},
% decode_binary(?TYPE_FLOAT, <<0.00012345678:32/float-little>>)),
%?assertEqual({1234.57, <<>>},
% decode_binary(?TYPE_FLOAT, <<1234.56789:32/float-little>>)),
ok.
null_bitmap_test() ->
?assertEqual({<<0, 1:1>>, <<>>}, null_bitmap_decode(9, <<0, 4>>, 2)),
?assertEqual(<<0, 4>>, null_bitmap_encode(<<0, 1:1>>, 2)),
ok.
lenenc_int_test() ->
%% decode
?assertEqual({40, <<>>}, lenenc_int(<<40>>)),
?assertEqual({16#ff, <<>>}, lenenc_int(<<16#fc, 255, 0>>)),
?assertEqual({16#33aaff, <<>>}, lenenc_int(<<16#fd, 16#ff, 16#aa, 16#33>>)),
?assertEqual({16#12345678, <<>>}, lenenc_int(<<16#fe, 16#78, 16#56, 16#34,
16#12, 0, 0, 0, 0>>)),
%% encode
?assertEqual(<<40>>, lenenc_int_encode(40)),
?assertEqual(<<16#fc, 255, 0>>, lenenc_int_encode(255)),
?assertEqual(<<16#fd, 16#ff, 16#aa, 16#33>>,
lenenc_int_encode(16#33aaff)),
?assertEqual(<<16#fe, 16#78, 16#56, 16#34, 16#12, 0, 0, 0, 0>>,
lenenc_int_encode(16#12345678)),
ok.
lenenc_str_test() ->
?assertEqual({<<"Foo">>, <<"bar">>}, lenenc_str(<<3, "Foobar">>)).
nulterm_test() ->
?assertEqual({<<"Foo">>, <<"bar">>}, nulterm_str(<<"Foo", 0, "bar">>)).
parse_header_test() ->
%% Example from "MySQL Internals", revision 307, section 14.1.3.3 EOF_Packet
Packet = <<16#05, 16#00, 16#00, 16#05, 16#fe, 16#00, 16#00, 16#02, 16#00>>,
<> = Packet,
%% Check header contents and body length
?assertEqual({size(Body), 5, false}, parse_packet_header(Header)),
ok.
add_packet_headers_test() ->
{Data, 43} = add_packet_headers(<<"foo">>, 42),
?assertEqual(<<3, 0, 0, 42, "foo">>, list_to_binary(Data)).
parse_ok_test() ->
Body = <<0, 5, 1, 2, 0, 0, 0, "Foo">>,
?assertEqual(#ok{affected_rows = 5,
insert_id = 1,
status = ?SERVER_STATUS_AUTOCOMMIT,
warning_count = 0,
msg = <<"Foo">>},
parse_ok_packet(Body)).
parse_error_test() ->
%% Protocol 4.1
Body = <<255, 42, 0, "#", "XYZxx", "Foo">>,
?assertEqual(#error{code = 42, state = <<"XYZxx">>, msg = <<"Foo">>},
parse_error_packet(Body)),
ok.
parse_eof_test() ->
%% Example from "MySQL Internals", revision 307, section 14.1.3.3 EOF_Packet
Packet = <<16#05, 16#00, 16#00, 16#05, 16#fe, 16#00, 16#00, 16#02, 16#00>>,
<<_Header:4/binary-unit:8, Body/binary>> = Packet,
%% Ignore header. Parse body as an eof_packet.
?assertEqual(#eof{warning_count = 0,
status = ?SERVER_STATUS_AUTOCOMMIT},
parse_eof_packet(Body)),
ok.
hash_password_test() ->
?assertEqual(<<222,207,222,139,41,181,202,13,191,241,
234,234,73,127,244,101,205,3,28,251>>,
hash_password(<<"foo">>, <<"abcdefghijklmnopqrst">>)).
-endif.