BitFlag.jl
provides an Enum
-like type for bit flag option values. The main
motivations are:
- Members have implicit numbering with incrementing powers of 2.
- Binary OR (
|
) and AND (&
) operations are supported among members. - Values are pretty-printed by name, with OR chains when multiple bits are set.
This implementation is a relatively minor modification of
Julia's Enum
type implementation.
To create a new BitFlag
type, use the @bitflag
macro, provide a name, an
optional integer type, and a list of the member options (and optional values).
A new definition can be given in inline form:
@bitflag BitFlagName[::BaseType] value1[=x] value2[=y]
or as a block definition:
@bitflag BitFlagName[::BaseType] begin
value1[=x]
value2[=y]
end
Automatic numbering starts at 1, but an initial flag value may be explicitly
set to the value of zero. If no explicit zero-valued member is given, then 0 is
not a valid value for the BitFlag
. In the following example, we build an
8-bit BitFlag
with no value for bit 3 (value of 4).
julia> @bitflag MyStyle::UInt8 S_NONE=0 S_BOLD S_ITALIC S_LARGE=8
Combinations can be made using standard binary operations:
julia> S_BOLD | S_LARGE
(S_LARGE | S_BOLD)::MyStyle = 0x09
julia> ans & S_ITALIC
S_NONE::MyStyle = 0x00
Conversion to and from integers is permitted, but only for valid combinations of values:
julia> Int(S_BOLD)
1
julia> Integer(S_ITALIC) # Abstract Integer uses native UInt8 type
0x02
julia> MyStyle(9)
(S_LARGE | S_BOLD)::MyStyle = 0x09
julia> MyStyle(4) # MyStyle does not have a flag at 4
ERROR: ArgumentError: invalid value for BitFlag MyStyle: 4
Stacktrace:
...
In the above examples, both the bit flag type and member instances are added to
the surrounding scope.
If some members have common or conflicting names — or if scoped names are
simply desired on principle — the @bitflagx
macro can be used instead.
This variation supports the same features and syntax as @bitflag
(with
respect to choosing the base integer type, inline versus block definitions,
and setting particular flag values), but the definitions are instead placed
within a [bare] module, avoiding adding anything but the module name to the
surrounding scope.
For example, the following avoids shadowing the sin
function:
julia> @bitflagx TrigFunctions sin cos tan csc sec cot
julia> TrigFunctions.sin
sin::TrigFunctions.T = 0x00000001
julia> sin(π)
0.0
julia> print(typeof(TrigFunctions.sin))
Main.TrigFunctions.T
Because the module is named TrigFunction
, the generated type must have
a different name.
By default, the name of the type is T
, but it may be overridden by choosing
using the keyword option T = new_name
as the first argument:
julia> @bitflagx T=type HyperbolicTrigFunctions sinh cosh tanh csch sech coth
julia> HyperbolicTrigFunctions.tanh
tanh::HyperbolicTrigFunctions.type = 0x00000004
julia> print(typeof(HyperbolicTrigFunctions.tanh))
Main.HyperbolicTrigFunctions.type
Each flag value is then printed with contextual information which is more user-friendly than a raw integer:
julia> S_BOLD
S_BOLD::MyStyle = 0x00000001
julia> S_BOLD | S_LARGE
(S_LARGE | S_BOLD)::MyStyle = 0x00000005
In a compact context (such as in multi-dimensional arrays), the pretty-printing takes on a shorter form:
julia> [S_NONE (S_BOLD | S_LARGE)]
1×2 Array{MyStyle,2}:
S_NONE S_LARGE|S_BOLD
julia> show(IOContext(stdout, :compact => true), S_BOLD | S_LARGE)
S_LARGE|S_BOLD
BitFlag
s support writing to and reading from streams as integers:
julia> io = IOBuffer();
julia> write(io, UInt8(9));
julia> seekstart(io);
julia> read(io, MyStyle)
(S_LARGE | S_BOLD)::MyStyle = 0x09