use substrate_data::Value;
use crate::case::Case;
pub(crate) fn cases() -> Vec<Case> {
let mut all = vec![
Case::new("null", "The null value: tag 0, no payload.").all_forms(Value::Null),
Case::new("bool_false", "false: tag 1.").all_forms(Value::Bool(false)),
Case::new("bool_true", "true: tag 2.").all_forms(Value::Bool(true)),
];
all.extend(ints());
all.extend(floats());
all.extend(non_finite());
all.extend(strings());
all.extend(text_numbers());
all
}
const INTS: [i64; 13] = [
0,
1,
-1,
63,
64,
-64,
-65,
8191,
8192,
2_147_483_647,
2_147_483_648,
i64::MAX,
i64::MIN,
];
fn ints() -> Vec<Case> {
INTS.iter()
.map(|&n| {
let name = if n < 0 {
format!("int_minus_{}", n.unsigned_abs())
} else {
format!("int_{n}")
};
let zigzag = ((n << 1) ^ (n >> 63)).cast_unsigned();
let case = Case::new(
&name,
&format!("The integer {n}: tag 3, then its zigzag form {zigzag} as a varint."),
);
case.all_forms(Value::Int(n))
})
.collect()
}
const FLOATS: [(&str, f64, &str); 12] = [
("zero", 0.0, "positive zero"),
(
"minus_zero",
-0.0,
"negative zero: its sign bit is kept in every form",
),
(
"one",
1.0,
"one, spelled 1.0 so the text reads back as a float",
),
("minus_one_and_a_half", -1.5, "a negative float"),
(
"tenth",
0.1,
"0.1, the shortest text that reads back as the same f64",
),
(
"half",
0.5,
"exactly an f32, still written as a full f64 (tag 4) in a plain value",
),
(
"ten_to_the_16",
1e16,
"text spells it 1e16, never 10000000000000000 (that reads as an int)",
),
("max", f64::MAX, "the largest finite f64"),
("min_positive", f64::MIN_POSITIVE, "the smallest normal f64"),
("subnormal", 5e-324, "the smallest subnormal f64"),
("small", 1e-7, "a small exponent"),
(
"f32_max",
3.402_823_466_385_288_6e38,
"the largest f32, as an f64",
),
];
fn floats() -> Vec<Case> {
FLOATS
.iter()
.map(|&(name, f, why)| {
let description =
format!("A float, {why}: tag 4, then the f64 as 8 bytes little-endian.");
Case::new(&format!("float_{name}"), &description).all_forms(Value::Float(f))
})
.collect()
}
fn non_finite() -> Vec<Case> {
let note = "Binary only: the text form writes a non-finite float as null (a different value), \
so no text is given. The binary form keeps the exact 64 bits, a NaN's payload and \
sign included; compare floats by their bits.";
let make = |name: &str, what: &str, bits: u64| {
let value = Value::Float(f64::from_bits(bits));
let bytes = substrate_data::binary::encode(&value);
Case::new(name, &format!("{what}. {note}"))
.value(value)
.binary(bytes)
};
vec![
make("float_infinity", "Positive infinity", 0x7ff0_0000_0000_0000),
make(
"float_minus_infinity",
"Negative infinity",
0xfff0_0000_0000_0000,
),
make("float_nan", "The quiet NaN", 0x7ff8_0000_0000_0000),
make(
"float_nan_payload",
"A NaN with a payload",
0x7ff8_0000_0000_0001,
),
make(
"float_nan_negative",
"A NaN with the sign bit set",
0xfff8_0000_0000_0000,
),
]
}
const PLAIN_STRINGS: [(&str, &str, &str); 11] = [
(
"empty",
"",
"The empty string: a new table entry of length 0.",
),
(
"ascii",
"hello",
"A short ASCII string: tag 5, a new-string marker (length << 1 | 1), the bytes.",
),
(
"latin",
"héllo wörld",
"Accented letters: UTF-8, the length counts bytes (13), not characters.",
),
("cjk", "日本語", "Three characters, nine bytes."),
(
"emoji",
"🦀 and 🧪",
"Characters outside the BMP: four bytes each in UTF-8.",
),
(
"escapes",
"quote \" backslash \\ newline \n tab \t",
"Text escapes these; binary holds them raw.",
),
(
"control",
"bell \u{7} and \u{1} and unit separator \u{1f}",
"Control characters: text writes them as \\u00xx.",
),
(
"separators",
"line \u{2028} and paragraph \u{2029} separators",
"Not escaped in text, kept raw.",
),
(
"at_doubled",
"@@everyone",
"A string as spelled: a plain text that starts with `@` is stored with it doubled.",
),
(
"looks_like_ref",
"@file:images/logo.png",
"A string spelled like a reference is still a string value in the model; only a schema reads it as a ref.",
),
(
"looks_like_bytes",
"@bytes:AQID",
"A string spelled like bytes is still a string value in the model.",
),
];
fn strings() -> Vec<Case> {
let long63 = "a".repeat(63);
let long64 = "a".repeat(64);
let mut out: Vec<Case> = PLAIN_STRINGS
.iter()
.map(|&(name, s, why)| Case::new(&format!("str_{name}"), why).all_forms(Value::from(s)))
.collect();
out.push(
Case::new(
"str_len_63",
"63 bytes: the marker 127 is the last one-byte varint.",
)
.all_forms(Value::from(long63)),
);
out.push(
Case::new("str_len_64", "64 bytes: the marker 129 takes two bytes.")
.all_forms(Value::from(long64)),
);
out
}
fn text_numbers() -> Vec<Case> {
let float = |name: &str, text: &str, f: f64, why: &str| {
Case::new(name, why).text(text).value(Value::Float(f))
};
vec![
float(
"text_exponent_is_float",
"1E5",
100_000.0,
"An exponent makes a float, even a whole one (the `E` may be upper case).",
),
float(
"text_exponent_two",
"1e2",
100.0,
"1e2 is the float 100.0, not the int 100.",
),
float(
"text_int_past_i64",
"9223372036854775808",
9.223_372_036_854_776e18,
"A whole number that does not fit i64 reads as a float.",
),
float(
"text_int_past_u64",
"18446744073709551616",
1.844_674_407_370_955_2e19,
"Likewise far past u64.",
),
float(
"text_float_overflows",
"1.5e400",
f64::INFINITY,
"A float too big for f64 reads as infinity (and writes back as null).",
),
Case::new(
"text_minus_zero_int",
"`-0` without a fraction is the int 0.",
)
.text("-0")
.value(Value::Int(0)),
Case::new("text_int_min", "i64::MIN reads as an int.")
.text("-9223372036854775808")
.value(Value::Int(i64::MIN)),
]
}