vo_gpu, options: don't return NaN through API

Internally, vo_gpu uses NaN for some options to indicate a default value
that is different depending on the context (e.g. different scalers).
There are 2 problems with this:

1. you couldn't reset the options to their defaults
2. NaN is a damn mess and shouldn't be part of the API

The option parser already rejected NaN explicitly, which is why 1.
didn't work. Regarding 2., JSON might be a good example, and actually
caused a bug report.

Fix this by mapping NaN to the special value "default". I think I'd
prefer other mechanisms (maybe just having every scaler expose separate
options?), but for now this will do. See you in a future commit, which
painfully deprecates this and replaces it with something else.

I refrained from using "no" (my favorite magic value for "unset" etc.)
because then I'd have e.g. make --no-scale-param1 work, which in
addition to a lot of effort looks dumb and nobody will use it.

Here's also an apology for the shitty added test script.

Fixes: #6691

DOCS/interface-changes.rst

@@ -191,6 +191,9 @@ Interface changes
       internal counter that remembered the current position.
     - remove deprecated ao/vo auto profiles. Consider using scripts like
       auto-profiles.lua instead.
+    - --[c]scale-[w]param[1|2] and --tone-mapping-param now accept "default",
+      and if set to that value, reading them as property will also return
+      "default", instead of float nan as in previous versions
  --- mpv 0.28.0 ---
     - rename --hwdec=mediacodec option to mediacodec-copy, to reflect
       conventions followed by other hardware video decoding APIs

DOCS/man/options.rst

@@ -4404,8 +4404,10 @@ The following video options are currently all specific to ``--vo=gpu`` and
     smooth.
 
 ``--scale-param1=<value>``, ``--scale-param2=<value>``, ``--cscale-param1=<value>``, ``--cscale-param2=<value>``, ``--dscale-param1=<value>``, ``--dscale-param2=<value>``, ``--tscale-param1=<value>``, ``--tscale-param2=<value>``
-    Set filter parameters. Ignored if the filter is not tunable. Currently,
+    Set filter parameters. By default, these are set to the special string
-    this affects the following filter parameters:
+    ``default``, which maps to a scaler-specific default value. Ignored if the
+    filter is not tunable. Currently, this affects the following filter
+    parameters:
 
     bcspline
         Spline parameters (``B`` and ``C``). Defaults to 0.5 for both.
@@ -4477,8 +4479,10 @@ The following video options are currently all specific to ``--vo=gpu`` and
 
 ``--scale-wparam=<window>``, ``--cscale-wparam=<window>``, ``--cscale-wparam=<window>``, ``--tscale-wparam=<window>``
     (Advanced users only) Configure the parameter for the window function given
-    by ``--scale-window`` etc. Ignored if the window is not tunable. Currently,
+    by ``--scale-window`` etc. By default, these are set to the special string
-    this affects the following window parameters:
+    ``default``, which maps to a window-specific default value. Ignored if the
+    window is not tunable. Currently, this affects the following window
+    parameters:
 
     kaiser
         Window parameter (alpha). Defaults to 6.33.
@@ -5524,8 +5528,10 @@ The following video options are currently all specific to ``--vo=gpu`` and
         the display.
 
 ``--tone-mapping-param=<value>``
-    Set tone mapping parameters. Ignored if the tone mapping algorithm is not
+    Set tone mapping parameters. By default, this is set to the special string
-    tunable. This affects the following tone mapping algorithms:
+    ``default``, which maps to an algorithm-specific default value. Ignored if
+    the tone mapping algorithm is not tunable. This affects the following tone
+    mapping algorithms:
 
     clip
         Specifies an extra linear coefficient to multiply into the signal

TOOLS/lua/nan-test.lua

@@ -0,0 +1,37 @@
+-- Test a float property which internally uses NaN.
+-- Run with --no-config (or just scale-param1 not set).
+
+local utils = require 'mp.utils'
+
+prop_name = "scale-param1"
+
+-- internal NaN, return string "default" instead of NaN
+v = mp.get_property_native(prop_name, "fail")
+print("Exp:", "string", "\"default\"")
+print("Got:", type(v), utils.to_string(v))
+
+v = mp.get_property(prop_name)
+print("Exp:", "default")
+print("Got:", v)
+
+-- not representable -> return provided fallback value
+v = mp.get_property_number(prop_name, -100)
+print("Exp:", -100)
+print("Got:", v)
+
+mp.set_property_native(prop_name, 123)
+v = mp.get_property_number(prop_name, -100)
+print("Exp:", "number", 123)
+print("Got:", type(v), utils.to_string(v))
+
+-- try to set an actual NaN
+st, msg = mp.set_property_number(prop_name, 0.0/0)
+print("Exp:", nil, "<message>")
+print("Got:", st, msg)
+
+-- set default
+mp.set_property(prop_name, "default")
+
+v = mp.get_property(prop_name)
+print("Exp:", "default")
+print("Got:", v)

options/m_option.c

@@ -918,6 +918,11 @@ static int parse_double(struct mp_log *log, const m_option_t *opt,
     if (bstr_eatstart0(&rest, ":") || bstr_eatstart0(&rest, "/"))
         tmp_float /= bstrtod(rest, &rest);
 
+    if ((opt->flags & M_OPT_DEFAULT_NAN) && bstr_equals0(param, "default")) {
+        tmp_float = NAN;
+        goto done;
+    }
+
     if (rest.len) {
         mp_err(log, "The %.*s option must be a floating point number or a "
                "ratio (numerator[:/]denominator): %.*s\n",
@@ -931,6 +936,7 @@ static int parse_double(struct mp_log *log, const m_option_t *opt,
         return M_OPT_OUT_OF_RANGE;
     }
 
+done:
     if (dst)
         VAL(dst) = tmp_float;
     return 1;
@@ -938,12 +944,18 @@ static int parse_double(struct mp_log *log, const m_option_t *opt,
 
 static char *print_double(const m_option_t *opt, const void *val)
 {
-    return talloc_asprintf(NULL, "%f", VAL(val));
+    double f = VAL(val);
+    if (isnan(f) && (opt->flags & M_OPT_DEFAULT_NAN))
+        return talloc_strdup(NULL, "default");
+    return talloc_asprintf(NULL, "%f", f);
 }
 
 static char *print_double_f3(const m_option_t *opt, const void *val)
 {
-    return talloc_asprintf(NULL, "%.3f", VAL(val));
+    double f = VAL(val);
+    if (isnan(f))
+        return print_double(opt, val);
+    return talloc_asprintf(NULL, "%.3f", f);
 }
 
 static void add_double(const m_option_t *opt, void *val, double add, bool wrap)
@@ -989,8 +1001,14 @@ static int double_set(const m_option_t *opt, void *dst, struct mpv_node *src)
 static int double_get(const m_option_t *opt, void *ta_parent,
                       struct mpv_node *dst, void *src)
 {
-    dst->format = MPV_FORMAT_DOUBLE;
+    double f = *(double *)src;
-    dst->u.double_ = *(double *)src;
+    if (isnan(f) && (opt->flags & M_OPT_DEFAULT_NAN)) {
+        dst->format = MPV_FORMAT_STRING;
+        dst->u.string = talloc_strdup(ta_parent, "default");
+    } else {
+        dst->format = MPV_FORMAT_DOUBLE;
+        dst->u.double_ = f;
+    }
     return 1;
 }
 
@@ -1023,12 +1041,14 @@ static int parse_float(struct mp_log *log, const m_option_t *opt,
 
 static char *print_float(const m_option_t *opt, const void *val)
 {
-    return talloc_asprintf(NULL, "%f", VAL(val));
+    double tmp = VAL(val);
+    return print_double(opt, &tmp);
 }
 
 static char *print_float_f3(const m_option_t *opt, const void *val)
 {
-    return talloc_asprintf(NULL, "%.3f", VAL(val));
+    double tmp = VAL(val);
+    return print_double_f3(opt, &tmp);
 }
 
 static void add_float(const m_option_t *opt, void *val, double add, bool wrap)
@@ -1057,9 +1077,8 @@ static int float_set(const m_option_t *opt, void *dst, struct mpv_node *src)
 static int float_get(const m_option_t *opt, void *ta_parent,
                      struct mpv_node *dst, void *src)
 {
-    dst->format = MPV_FORMAT_DOUBLE;
+    double tmp = VAL(src);
-    dst->u.double_ = VAL(src);
+    return double_get(opt, ta_parent, dst, &tmp);
-    return 1;
 }
 
 const m_option_type_t m_option_type_float = {

options/m_option.h

@@ -424,6 +424,9 @@ char *format_file_size(int64_t size);
 #define UPDATE_OPTS_MASK \
     (((UPDATE_OPT_LAST << 1) - 1) & ~(unsigned)(UPDATE_OPT_FIRST - 1))
 
+// type_float/type_double: string "default" is parsed as NaN (and reverse)
+#define M_OPT_DEFAULT_NAN       (1 << 29)
+
 // Like M_OPT_TYPE_OPTIONAL_PARAM.
 #define M_OPT_OPTIONAL_PARAM    (1 << 30)
 

video/out/gpu/video.c

@@ -342,14 +342,18 @@ static int validate_error_diffusion_opt(struct mp_log *log, const m_option_t *op
 
 #define OPT_BASE_STRUCT struct gl_video_opts
 
+// Use for options which use NAN for defaults.
+#define OPT_FLOATDEF(name, var, flags) \
+    OPT_FLOAT(name, var, (flags) | M_OPT_DEFAULT_NAN)
+
 #define SCALER_OPTS(n, i) \
     OPT_STRING_VALIDATE(n, scaler[i].kernel.name, 0, validate_scaler_opt), \
-    OPT_FLOAT(n"-param1", scaler[i].kernel.params[0], 0),                  \
+    OPT_FLOATDEF(n"-param1", scaler[i].kernel.params[0], 0),               \
-    OPT_FLOAT(n"-param2", scaler[i].kernel.params[1], 0),                  \
+    OPT_FLOATDEF(n"-param2", scaler[i].kernel.params[1], 0),               \
     OPT_FLOAT(n"-blur",   scaler[i].kernel.blur, 0),                       \
     OPT_FLOATRANGE(n"-cutoff", scaler[i].cutoff, 0, 0.0, 1.0),             \
     OPT_FLOATRANGE(n"-taper", scaler[i].kernel.taper, 0, 0.0, 1.0),        \
-    OPT_FLOAT(n"-wparam", scaler[i].window.params[0], 0),                  \
+    OPT_FLOATDEF(n"-wparam", scaler[i].window.params[0], 0),               \
     OPT_FLOAT(n"-wblur",  scaler[i].window.blur, 0),                       \
     OPT_FLOATRANGE(n"-wtaper", scaler[i].window.taper, 0, 0.0, 1.0),       \
     OPT_FLOATRANGE(n"-clamp", scaler[i].clamp, 0, 0.0, 1.0),               \
@@ -383,7 +387,7 @@ const struct m_sub_options gl_video_conf = {
                        tone_map.scene_threshold_low, 0, 0, 20.0),
         OPT_FLOATRANGE("hdr-scene-threshold-high",
                        tone_map.scene_threshold_high, 0, 0, 20.0),
-        OPT_FLOAT("tone-mapping-param", tone_map.curve_param, 0),
+        OPT_FLOATDEF("tone-mapping-param", tone_map.curve_param, 0),
         OPT_FLOATRANGE("tone-mapping-max-boost", tone_map.max_boost, 0, 1.0, 10.0),
         OPT_FLOAT("tone-mapping-desaturate", tone_map.desat, 0),
         OPT_FLOATRANGE("tone-mapping-desaturate-exponent",