On enums and scope

[russel]

With the libdvbv5 C header file transform, a lot of named enums are created. This is fine per se. So for example:

typedef enum fe_delivery_system {
	SYS_UNDEFINED,
	SYS_DVBC_ANNEX_A,
	SYS_DVBC_ANNEX_B,
	SYS_DVBT,
	SYS_DSS,
	SYS_DVBS,
	SYS_DVBS2,
	SYS_DVBH,
	SYS_ISDBT,
	SYS_ISDBS,
	SYS_ISDBC,
	SYS_ATSC,
	SYS_ATSCMH,
	SYS_DTMB,
	SYS_CMMB,
	SYS_DAB,
	SYS_DVBT2,
	SYS_TURBO,
	SYS_DVBC_ANNEX_C,
} fe_delivery_system_t;

is transformed to:

enum fe_delivery_system
{
    SYS_UNDEFINED = 0,
    SYS_DVBC_ANNEX_A = 1,
    SYS_DVBC_ANNEX_B = 2,
    SYS_DVBT = 3,
    SYS_DSS = 4,
    SYS_DVBS = 5,
    SYS_DVBS2 = 6,
    SYS_DVBH = 7,
    SYS_ISDBT = 8,
    SYS_ISDBS = 9,
    SYS_ISDBC = 10,
    SYS_ATSC = 11,
    SYS_ATSCMH = 12,
    SYS_DTMB = 13,
    SYS_CMMB = 14,
    SYS_DAB = 15,
    SYS_DVBT2 = 16,
    SYS_TURBO = 17,
    SYS_DVBC_ANNEX_C = 18
}

alias fe_delivery_system_t = fe_delivery_system;

However, the line:

#define SYS_DVBC_ANNEX_AC	SYS_DVBC_ANNEX_A

gets transformed into:

enum SYS_DVBC_ANNEX_AC = SYS_DVBC_ANNEX_A;

where it needs to be transformed to:

enum SYS_DVBC_ANNEX_AC = fe_delivery_system.SYS_DVBC_ANNEX_A;

[jacob-carlborg]

I'm not sure if this is possible. When the compiler is analyzing #define SYS_DVBC_ANNEX_AC SYS_DVBC_ANNEX_A, I'm pretty sure it has no idea that SYS_DVBC_ANNEX_A is the same as the one defined in fe_delivery_system. The only way I can see this working is if DStep added aliases for all enum members to expose them outside the enum.

[russel]

Hummm… C enum symbols are global and must be named appropriately to avoid clashes. D scopes the symbols and is better for that.

I guess the only way is to keep a temporary look up table of all enum symbols and then do a look up to impose the scope. This may be more effort that it is worth.

[jacob-carlborg]

Hummm… C enum symbols are global and must be named appropriately to avoid clashes. D scopes the symbols and is better for that.

I think it would be reasonable to have a flag to add aliases for all enum members. That would allow to generate bindings with a closer API to the C API.

I guess the only way is to keep a temporary look up table of all enum symbols and then do a look up to impose the scope. This may be more effort that it is worth.

Perhaps that could work.