NML - a Newgrf Meta Language

[Hirundo]

And... did you try with "the latest" nightly (yesterday / today)?

It seems my local 'trunk' copy wasn't too up to date either

[planetmaker]

Sorry to cause all this fuzz about something that already was fixed, but thanks very much for your time trying to help me out here (and eventually helping me out)!

No worries. I remembered dimmly that commit, so I thought I just ask

[Yexo]

Since grf v8 I think there is no reason anymore to hardcode prop 15 to 0xFF, it can now actually be a default cargo. As such it's a bug in NML as well (although that was not what you experienced) .

[FooBar]

it can now actually be a default cargo

But what would you set it to in my case?
It's a passengers-only train, but I don't have passengers in my cargotable.

Or are you saying that NML should allow setting a default cargo by the coder (or at their discretion, the first refittable), basically what akasoft wanted to do a number of posts back?

[akasoft]

If I use code

Code: Select all

switch (FEAT_TRAINS, SELF, _2m62u_cargo_subtype_text2,
  current_year >= 1990)
{
  1: string(STR_REFIT_V3);
  return CB_RESULT_NO_TEXT;
}

switch (FEAT_TRAINS, SELF, _2m62u_cargo_subtype_text,
  cargo_subtype)
{
  0: string(STR_REFIT_V1);
  1: string(STR_REFIT_V2);
  2: _2m62u_cargo_subtype_text2;
  3: string(STR_REFIT_V4);
  4: string(STR_REFIT_V5);
  5: string(STR_REFIT_V6);
  6: string(STR_REFIT_V7);
  7: string(STR_REFIT_V8);
  8: string(STR_REFIT_V9);
  return CB_RESULT_NO_TEXT;
}

it work normally, but if I try to simplify the code so

Code: Select all

switch (FEAT_TRAINS, SELF, _2m62u_cargo_subtype_text,
  cargo_subtype)
{
  0: string(STR_REFIT_V1);
  1: string(STR_REFIT_V2);
  2: return ((current_year >= 1990) ? string(STR_REFIT_V3) : CB_RESULT_NO_TEXT);
  3: string(STR_REFIT_V4);
  4: string(STR_REFIT_V5);
  5: string(STR_REFIT_V6);
  6: string(STR_REFIT_V7);
  7: string(STR_REFIT_V8);
  8: string(STR_REFIT_V9);
  return CB_RESULT_NO_TEXT;
}

then I receive the error when try to refit cargo

cb19_error.png (2.95 KiB) Viewed 3256 times

Why, what's wrong? I cannot use syntax "() ? () : ()" here?

[planetmaker]

(...)Why, what's wrong? I cannot use syntax "() ? () : ()" here?

While I agree that it would be nice, if it works, the NML manual doesn't tell that it is correct: http://newgrf-specs.tt-wiki.net/wiki/NML:Switch

what I'd try is to use temporary registers:

Code: Select all

switch (FEAT_TRAINS, SELF, _2m62u_cargo_subtype_text,
  [STORE_TEMP( (current_year >= 1990) ? string(STR_REFIT_V3) : CB_RESULT_NO_TEXT, 0), cargo_subtype])
{
  0: string(STR_REFIT_V1);
  1: string(STR_REFIT_V2);
  2: return LOAD_TEMP(0);
  3: string(STR_REFIT_V4);
  4: string(STR_REFIT_V5);
  5: string(STR_REFIT_V6);
  6: string(STR_REFIT_V7);
  7: string(STR_REFIT_V8);
  8: string(STR_REFIT_V9);
  return CB_RESULT_NO_TEXT;
}

Note that I haven't tested it.

Note also, that you should always use 'return string(...)' instead of just 'string(...)'. The way you write without return does work indeed, but it's undocumented and might cease to work without notice (yes, I'm a bad example there, too. My newgrfs have that code without return as well )

[akasoft]

I thought that "expression" include "()?():()".

Thank you for the example of the temporary registers. And notes.

I try to use this hack (considering that CB_RESULT_NO_TEXT == 0x400)

Code: Select all

  2: return ((current_year >= 1990) ? string(STR_REFIT_V3) : CB_RESULT_NO_TEXT) & 0x7FF;

it work normally. This is malicious hack or not?

Variant of the temporary registers not work and receive the same error message.
I think, the key point here - "string(..)" and "& 0x7FF". Function string(..) has the return value with setted high bits. Or their set the "()?():()". Or "return" not work properly.

My version of "()?():()" I like more than.

[planetmaker]

I thought that "expression" include "()?():()".

You're right.

Code: Select all

  2: return ((current_year >= 1990) ? string(STR_REFIT_V3) : CB_RESULT_NO_TEXT) & 0x7FF;

it work normally. This is malicious hack or not?

Malicious certainly not. But it looks wrong to me... on first look the &0x7FF imho shouldn't be required.

[akasoft]

Add.
Code

Code: Select all

switch (FEAT_TRAINS, SELF, _2m62u_cargo_subtype_text,
[
  STORE_TEMP(string(STR_REFIT_V3), 0),
  cargo_subtype
]
)
{
  0: string(STR_REFIT_V1);
  1: string(STR_REFIT_V2);
  2: return LOAD_TEMP(0);
  3: string(STR_REFIT_V4);
  4: string(STR_REFIT_V5);
  5: string(STR_REFIT_V6);
  6: string(STR_REFIT_V7);
  7: string(STR_REFIT_V8);
  8: string(STR_REFIT_V9);
  return CB_RESULT_NO_TEXT;
}

also not work and receive the error message.

[akasoft]

If used

Code: Select all

  STORE_TEMP(string(STR_REFIT_V3) & 0x7FF, 0),

then has no error and it work.

nmlc.exe r1792
openttd r23840 trunk x86
Windows 7 x64

[Hirundo]

[quote="akasoft"]

Code: Select all

 return ((current_year >= 1990) ? string(STR_REFIT_V3) : CB_RESULT_NO_TEXT);

The () ? () : () operator is designed to operate on numeric values only. A string() is not a number, although it NML tries to make sense of it, it fails quite badly.

For now, don't use ?: on strings. In the future, NML should be amended to either properly handle this case or throw an error. I have filed an NML ticket (3610) about this

[akasoft]

For now, don't use ?: on strings.

Why? First, it is very convenient, is easy to do. Secondly, the function string () returns a number.
In terms of the openttd, it has no strings, only numbers.

Please, do not disable this feature. There's just need to make the return value into the valid value. Which I did with using "& 0x7FF".

Add: At NML (NewGRF) are not so many opportunities for programming, so they throw away.

[akasoft]

How to use procedure call in NML?

If I use code:

Code: Select all

switch (FEAT_TRAINS, SELF, l_e_purchase_speed,
  current_year >= 1920)
{
  1: return 120;
  return 105;
}

switch (FEAT_TRAINS, SELF, l_e_speed,
[
  STORE_TEMP(var[0x7E, 0, 0, l_e_purchase_speed] + last_computed_result, 0),
  position_in_consist == 0 ? LOAD_TEMP(0) : LOAD_TEMP(0) * 85 / 100
])
{
  return;
}

then receive nmlc error: "nmlc: "xussr.nml", line 1263: All parts of a variable access must be integers.".

I want to use block l_e_purchase_speed via procedure call.
What is the correct syntax to use it?

[Yexo]

What is the correct syntax to use it?

There is no syntax for it yet, it's currently unsupported.

[akasoft]

It will be supported? Is planned?

[Yexo]

Yes, but it has been that way for a while now. Don't count on it anytime soon.

[oberhümer]

Is is possible to force a 32 bpp blitter (or "set it as preferred") using NML?

[FooBar]

You can never force that, also not in NFO. What this does is request the game for a 32bpp blitter, but there's no guarantee that you'll get one.

Either way, NML sets the appropriate action 14 entry automatically if you define 32bpp sprites in your NML code. See http://dev.openttdcoop.org/projects/nml ... 78ee234135

[FooBar]

Sorry for the doublepost, but I was wondering what the best approach is for livery refits on an engine that isn't supposed to have a capacity.

I guess I'm supposed to give it a capacity and then override that again with the cargo_capacity callback, but I can't quite figure it out how it's all supposed to come together.

[frosch]

Yes, you need to set the property to a non-zero value, and set a zero capacity with the callback.
You will also have to decide for some cargo type.