Add minetest.get_player_window_information()

[rubenwardy]

\Adds a new packet to send dynamic client info to the server, and minetest.get_player_window_information()

Fixes #10632

* `minetest.get_player_window_information(player_name)`:

      -- Will only be present if the client sent this information (requires v5.7+)
      --
      -- Note that none of these things are constant, they are likely to change during a client
      -- connection as the player resizes the window and moves it between monitors
      --
      -- real_gui_scaling and real_hud_scaling can be used instead of DPI.
      -- OSes don't necessarily give the physical DPI, as they may allow user configuration.
      -- real_*_scaling is just OS DPI / 96 but with another level of user configuration.
      {
          -- Current size of the in-game render target.
          --
          -- This is usually the window size, but may be smaller in certain situations,
          -- such as side-by-side mode.
          size = {
              x = 1308,
              y = 577,
          },

          -- Estimated maximum formspec size before Minetest will start shrinking the
          -- formspec to fit. For a fullscreen formspec, use a size 10-20% larger than
          -- this and `padding[-0.01,-0.01]`.
          max_formspec_size = {
              x = 20,
              y = 11.25
          },

          -- GUI Scaling multiplier
          -- Equal to the setting `gui_scaling` multiplied by `dpi / 96`
          real_gui_scaling = 1,

          -- HUD Scaling multiplier
          -- Equal to the setting `hud_scaling` multiplied by `dpi / 96`
          real_hud_scaling = 1,
      }

To do

This PR is ready for review

• documentation
• Testing
• Return display = nil when the client doesn't send info
• DPI is always an integer on Windows. Is it always an integer on other platforms?
• should this be a new API function?
• debounce sending
• What other display information is needed?
• Resist fingerprinting
• Move to player ObjectRef?
• Add note that real_gui_scaling/real_hud_scaling can be used instead of DPI

How to test

• See /testfullscreenfs
• Test in sidebyside 3d mode

[MisterE123]

wow, what an amazingly great and awsome thing.

[sfan5]

    dpi = 96, -- Dots per inch
    gui_scaling = 1, -- gui_scaling setting
    hud_scaling = 1, -- hud_scaling setting

Any reason to separate these?
Having real_hud_scaling & real_gui_scaling would convey the necessary info without mods needing to do math every time.
(real_hud_scaling := hud_scaling * dpi / 96 in case that isn't obvious)

    screen_size = { -- Current window size, not including any decorations (window bar, etc)
        x = 1308,
        y = 577,
    },

1. This can potentially be minimized into an approximate aspect ratio plus height
2. What's the usecase here? Sizing of GUI elements?
3. The description should say "current size of the in-game area" or similar to be future-proof.

should this be a new API function?

Definitely, "you call the API and get 20 values of which you need only two" is already an issue with get_player_information. We shouldn't make it worse.

DPI is always an integer on Windows. Is it always an integer on other platforms?

Considering the above formula the fractional part of DPI will change the scale of things by less than ~0.01, so rounding to integer is safe to do.

What other display information is needed?

Current enablement state of the touchscreen GUI.

[MisterE123]

@sfan5, yes the usecase is sizing elements. with this information, it is possible to make fullscreen formspecs! I have made fullscreen formspecs in the mainmenu using this information (which passes it) , and it works pretty well.

also, its possible to just make sure that elements fit well on the screen.

[rubenwardy]

    dpi = 96, -- Dots per inch
    gui_scaling = 1, -- gui_scaling setting
    hud_scaling = 1, -- hud_scaling setting

Any reason to separate these? Having real_hud_scaling & real_gui_scaling would convey the necessary info without mods needing to do math every time. (real_hud_scaling := hud_scaling * dpi / 96 in case that isn't obvious)

Looking into it, yeah it would make sense to combine these. Raw DPI is only useful if you want to know the physical size of the screen, but this dpi value can be overridden by screen_dpi anyway

1. This can potentially be minimized into an approximate aspect ratio plus height

This seems a bit derived to me, It's not minimized, it's the same number of numbers just a derived value rather than the raw values

2. What's the usecase here? Sizing of GUI elements?

The use case is knowing how much space is available so that you can 1) make a formspec fullsize 2) make the best use of space

If a modder wants to make a formspec full size, then they just need the ratio to calculate a large size[] with the same ratio as the window. Minetest will scale down formspec windows to fit the game window. But having the additional information allows for simplifying or adding more to the UI when there is more space

3. The description should say "current size of the in-game area" or similar to be future-proof.

Makes sense, side-by-side mode means that only half of the window's width is used

DPI is always an integer on Windows. Is it always an integer on other platforms?

Considering the above formula the fractional part of DPI will change the scale of things by less than ~0.01, so rounding to integer is safe to do.

guiFormspecMenu uses it as a double, which is a bit overkill. Given that we've decided to combine this into the _scaling settings, there's no need for a change here

What other display information is needed?

Current enablement state of the touchscreen GUI.

See #12264. This should be sent as part of the ClientDynamicInfo packet, but I don't know what the Lua API should be

[rubenwardy]

A value that may be useful would be something like this:

-- Max formspec size before Minetest starts scaling down, assuming 
-- `fixed_size` is true in `size[]` and`padding[0,0]`.
--
-- This can be used to make fullscreen formspecs by using a padding
-- of `padding[-0.01,-0.01]` and a small inset `0.1` between the edge
-- of the formspec and the elements.
max_formspec_size = {
    x = 10.3,
    y = 5.4,
}

[rubenwardy]

Added a mod to devtest to test fullscreen formspecs

[rubenwardy]

A value that may be useful would be something like this:

-- Max formspec size before Minetest starts scaling down, assuming 
-- `fixed_size` is true in `size[]` and`padding[0,0]`.
--
-- This can be used to make fullscreen formspecs by using a padding
-- of `padding[-0.01,-0.01]` and a small inset `0.1` between the edge
-- of the formspec and the elements.
max_formspec_size = {
    x = 10.3,
    y = 5.4,
}

This value is calculated by:

local function calculate_max_fs_size(name)
	local window = minetest.get_player_window_information(name)
	local slot_size = 0.5555 * 96 * window.real_gui_scaling
	return {
		x = window.size.x / slot_size,
		y = window.size.y / slot_size,
	}
end

This 0.5555 is taken from guiFormspecMenu.cpp: https://github.com/minetest/minetest/blob/master/src/gui/guiFormSpecMenu.cpp#L3216-L3223

This is only the case when true is provided to size[x,y;true]

[appgurueu]

For the API it'd be useful if this was event-driven rather than a getter (or probably both) - as a modder I prefer not to poll.

[rubenwardy]

That can come later

[Zughy]

@rubenwardy rebase needed

[sfan5]

Works.
my test code:

local thing = {}

minetest.register_on_joinplayer(function(player)
	local name = player:get_player_name()
	thing[name] = {
		hud = player:hud_add({
			hud_elem_type = "text",
			position = {x = 0.22, y = 0.22},
			alignment = {x = 1, y = 0},
			scale = {x = 2, y = 2},
			number = 0xFFFFFF,
			style = 4,
		}),
		text = ""
	}
end)

minetest.register_on_leaveplayer(function(player)
	thing[player:get_player_name()] = nil
end)

minetest.register_globalstep(function(dtime)
	for _, player in ipairs(minetest.get_connected_players()) do
		local name = player:get_player_name()
		local c = thing[name]
		if c then
			local s = dump(minetest.get_player_window_information(name))
			if s ~= c.text then
				player:hud_change(c.hud, "text", s)
				c.text = s
			end
		end
	end
end)