mirror of
https://github.com/denoland/deno.git
synced 2024-11-04 08:54:20 -05:00
9837d324a7
and update to prettier 1.19 Also, update `assert()` and remove not null assertions where possibly in `cli`. Closes #3273
668 lines
24 KiB
TypeScript
668 lines
24 KiB
TypeScript
import { sprintf } from "./sprintf.ts";
|
||
|
||
import { assertEquals } from "../testing/asserts.ts";
|
||
import { test, runIfMain } from "../testing/mod.ts";
|
||
|
||
const S = sprintf;
|
||
|
||
test(function noVerb(): void {
|
||
assertEquals(sprintf("bla"), "bla");
|
||
});
|
||
|
||
test(function percent(): void {
|
||
assertEquals(sprintf("%%"), "%");
|
||
assertEquals(sprintf("!%%!"), "!%!");
|
||
assertEquals(sprintf("!%%"), "!%");
|
||
assertEquals(sprintf("%%!"), "%!");
|
||
});
|
||
test(function testBoolean(): void {
|
||
assertEquals(sprintf("%t", true), "true");
|
||
assertEquals(sprintf("%10t", true), " true");
|
||
assertEquals(sprintf("%-10t", false), "false ");
|
||
assertEquals(sprintf("%t", false), "false");
|
||
assertEquals(sprintf("bla%t", true), "blatrue");
|
||
assertEquals(sprintf("%tbla", false), "falsebla");
|
||
});
|
||
|
||
test(function testIntegerB(): void {
|
||
assertEquals(S("%b", 4), "100");
|
||
assertEquals(S("%b", -4), "-100");
|
||
assertEquals(
|
||
S("%b", 4.1),
|
||
"100.0001100110011001100110011001100110011001100110011"
|
||
);
|
||
assertEquals(
|
||
S("%b", -4.1),
|
||
"-100.0001100110011001100110011001100110011001100110011"
|
||
);
|
||
assertEquals(
|
||
S("%b", Number.MAX_SAFE_INTEGER),
|
||
"11111111111111111111111111111111111111111111111111111"
|
||
);
|
||
assertEquals(
|
||
S("%b", Number.MIN_SAFE_INTEGER),
|
||
"-11111111111111111111111111111111111111111111111111111"
|
||
);
|
||
// width
|
||
|
||
assertEquals(S("%4b", 4), " 100");
|
||
});
|
||
|
||
test(function testIntegerC(): void {
|
||
assertEquals(S("%c", 0x31), "1");
|
||
assertEquals(S("%c%b", 0x31, 1), "11");
|
||
assertEquals(S("%c", 0x1f4a9), "💩");
|
||
//width
|
||
assertEquals(S("%4c", 0x31), " 1");
|
||
});
|
||
|
||
test(function testIntegerD(): void {
|
||
assertEquals(S("%d", 4), "4");
|
||
assertEquals(S("%d", -4), "-4");
|
||
assertEquals(S("%d", Number.MAX_SAFE_INTEGER), "9007199254740991");
|
||
assertEquals(S("%d", Number.MIN_SAFE_INTEGER), "-9007199254740991");
|
||
});
|
||
|
||
test(function testIntegerO(): void {
|
||
assertEquals(S("%o", 4), "4");
|
||
assertEquals(S("%o", -4), "-4");
|
||
assertEquals(S("%o", 9), "11");
|
||
assertEquals(S("%o", -9), "-11");
|
||
assertEquals(S("%o", Number.MAX_SAFE_INTEGER), "377777777777777777");
|
||
assertEquals(S("%o", Number.MIN_SAFE_INTEGER), "-377777777777777777");
|
||
// width
|
||
assertEquals(S("%4o", 4), " 4");
|
||
});
|
||
test(function testIntegerx(): void {
|
||
assertEquals(S("%x", 4), "4");
|
||
assertEquals(S("%x", -4), "-4");
|
||
assertEquals(S("%x", 9), "9");
|
||
assertEquals(S("%x", -9), "-9");
|
||
assertEquals(S("%x", Number.MAX_SAFE_INTEGER), "1fffffffffffff");
|
||
assertEquals(S("%x", Number.MIN_SAFE_INTEGER), "-1fffffffffffff");
|
||
// width
|
||
assertEquals(S("%4x", -4), " -4");
|
||
assertEquals(S("%-4x", -4), "-4 ");
|
||
// plus
|
||
assertEquals(S("%+4x", 4), " +4");
|
||
assertEquals(S("%-+4x", 4), "+4 ");
|
||
});
|
||
test(function testIntegerX(): void {
|
||
assertEquals(S("%X", 4), "4");
|
||
assertEquals(S("%X", -4), "-4");
|
||
assertEquals(S("%X", 9), "9");
|
||
assertEquals(S("%X", -9), "-9");
|
||
assertEquals(S("%X", Number.MAX_SAFE_INTEGER), "1FFFFFFFFFFFFF");
|
||
assertEquals(S("%X", Number.MIN_SAFE_INTEGER), "-1FFFFFFFFFFFFF");
|
||
});
|
||
|
||
test(function testFloate(): void {
|
||
assertEquals(S("%e", 4), "4.000000e+00");
|
||
assertEquals(S("%e", -4), "-4.000000e+00");
|
||
assertEquals(S("%e", 4.1), "4.100000e+00");
|
||
assertEquals(S("%e", -4.1), "-4.100000e+00");
|
||
assertEquals(S("%e", Number.MAX_SAFE_INTEGER), "9.007199e+15");
|
||
assertEquals(S("%e", Number.MIN_SAFE_INTEGER), "-9.007199e+15");
|
||
});
|
||
test(function testFloatE(): void {
|
||
assertEquals(S("%E", 4), "4.000000E+00");
|
||
assertEquals(S("%E", -4), "-4.000000E+00");
|
||
assertEquals(S("%E", 4.1), "4.100000E+00");
|
||
assertEquals(S("%E", -4.1), "-4.100000E+00");
|
||
assertEquals(S("%E", Number.MAX_SAFE_INTEGER), "9.007199E+15");
|
||
assertEquals(S("%E", Number.MIN_SAFE_INTEGER), "-9.007199E+15");
|
||
assertEquals(S("%E", Number.MIN_VALUE), "5.000000E-324");
|
||
assertEquals(S("%E", Number.MAX_VALUE), "1.797693E+308");
|
||
});
|
||
test(function testFloatfF(): void {
|
||
assertEquals(S("%f", 4), "4.000000");
|
||
assertEquals(S("%F", 4), "4.000000");
|
||
assertEquals(S("%f", -4), "-4.000000");
|
||
assertEquals(S("%F", -4), "-4.000000");
|
||
assertEquals(S("%f", 4.1), "4.100000");
|
||
assertEquals(S("%F", 4.1), "4.100000");
|
||
assertEquals(S("%f", -4.1), "-4.100000");
|
||
assertEquals(S("%F", -4.1), "-4.100000");
|
||
assertEquals(S("%f", Number.MAX_SAFE_INTEGER), "9007199254740991.000000");
|
||
assertEquals(S("%F", Number.MAX_SAFE_INTEGER), "9007199254740991.000000");
|
||
assertEquals(S("%f", Number.MIN_SAFE_INTEGER), "-9007199254740991.000000");
|
||
assertEquals(S("%F", Number.MIN_SAFE_INTEGER), "-9007199254740991.000000");
|
||
assertEquals(S("%f", Number.MIN_VALUE), "0.000000");
|
||
assertEquals(
|
||
S("%.324f", Number.MIN_VALUE),
|
||
// eslint-disable-next-line max-len
|
||
"0.000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000005"
|
||
);
|
||
assertEquals(S("%F", Number.MIN_VALUE), "0.000000");
|
||
assertEquals(
|
||
S("%f", Number.MAX_VALUE),
|
||
// eslint-disable-next-line max-len
|
||
"179769313486231570000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000.000000"
|
||
);
|
||
assertEquals(
|
||
S("%F", Number.MAX_VALUE),
|
||
// eslint-disable-next-line max-len
|
||
"179769313486231570000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000.000000"
|
||
);
|
||
});
|
||
|
||
test(function testString(): void {
|
||
assertEquals(S("%s World%s", "Hello", "!"), "Hello World!");
|
||
});
|
||
|
||
test(function testHex(): void {
|
||
assertEquals(S("%x", "123"), "313233");
|
||
assertEquals(S("%x", "n"), "6e");
|
||
});
|
||
test(function testHeX(): void {
|
||
assertEquals(S("%X", "123"), "313233");
|
||
assertEquals(S("%X", "n"), "6E");
|
||
});
|
||
|
||
test(function testType(): void {
|
||
assertEquals(S("%T", new Date()), "object");
|
||
assertEquals(S("%T", 123), "number");
|
||
assertEquals(S("%T", "123"), "string");
|
||
assertEquals(S("%.3T", "123"), "str");
|
||
});
|
||
|
||
test(function testPositional(): void {
|
||
assertEquals(S("%[1]d%[2]d", 1, 2), "12");
|
||
assertEquals(S("%[2]d%[1]d", 1, 2), "21");
|
||
});
|
||
|
||
test(function testSharp(): void {
|
||
assertEquals(S("%#x", "123"), "0x313233");
|
||
assertEquals(S("%#X", "123"), "0X313233");
|
||
assertEquals(S("%#x", 123), "0x7b");
|
||
assertEquals(S("%#X", 123), "0X7B");
|
||
assertEquals(S("%#o", 123), "0173");
|
||
assertEquals(S("%#b", 4), "0b100");
|
||
});
|
||
|
||
test(function testWidthAndPrecision(): void {
|
||
assertEquals(
|
||
S("%9.99d", 9),
|
||
// eslint-disable-next-line max-len
|
||
"000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000009"
|
||
);
|
||
assertEquals(S("%1.12d", 9), "000000000009");
|
||
assertEquals(S("%2s", "a"), " a");
|
||
assertEquals(S("%2d", 1), " 1");
|
||
assertEquals(S("%#4x", 1), " 0x1");
|
||
|
||
assertEquals(
|
||
S("%*.99d", 9, 9),
|
||
// eslint-disable-next-line max-len
|
||
"000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000009"
|
||
);
|
||
assertEquals(
|
||
S("%9.*d", 99, 9),
|
||
// eslint-disable-next-line max-len
|
||
"000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000009"
|
||
);
|
||
assertEquals(S("%*s", 2, "a"), " a");
|
||
assertEquals(S("%*d", 2, 1), " 1");
|
||
assertEquals(S("%#*x", 4, 1), " 0x1");
|
||
});
|
||
|
||
test(function testDash(): void {
|
||
assertEquals(S("%-2s", "a"), "a ");
|
||
assertEquals(S("%-2d", 1), "1 ");
|
||
});
|
||
test(function testPlus(): void {
|
||
assertEquals(S("%-+3d", 1), "+1 ");
|
||
assertEquals(S("%+3d", 1), " +1");
|
||
assertEquals(S("%+3d", -1), " -1");
|
||
});
|
||
|
||
test(function testSpace(): void {
|
||
assertEquals(S("% -3d", 3), " 3 ");
|
||
});
|
||
|
||
test(function testZero(): void {
|
||
assertEquals(S("%04s", "a"), "000a");
|
||
});
|
||
|
||
// relevant test cases from fmt_test.go
|
||
// eslint-disable-next-line @typescript-eslint/no-explicit-any
|
||
const tests: Array<[string, any, string]> = [
|
||
["%d", 12345, "12345"],
|
||
["%v", 12345, "12345"],
|
||
["%t", true, "true"],
|
||
|
||
// basic string
|
||
["%s", "abc", "abc"],
|
||
// ["%q", "abc", `"abc"`], // TODO: need %q?
|
||
["%x", "abc", "616263"],
|
||
["%x", "\xff\xf0\x0f\xff", "fff00fff"],
|
||
["%X", "\xff\xf0\x0f\xff", "FFF00FFF"],
|
||
["%x", "", ""],
|
||
["% x", "", ""],
|
||
["%#x", "", ""],
|
||
["%# x", "", ""],
|
||
["%x", "xyz", "78797a"],
|
||
["%X", "xyz", "78797A"],
|
||
["% x", "xyz", "78 79 7a"],
|
||
["% X", "xyz", "78 79 7A"],
|
||
["%#x", "xyz", "0x78797a"],
|
||
["%#X", "xyz", "0X78797A"],
|
||
["%# x", "xyz", "0x78 0x79 0x7a"],
|
||
["%# X", "xyz", "0X78 0X79 0X7A"],
|
||
|
||
// basic bytes : TODO special handling for Buffer? other std types?
|
||
// escaped strings : TODO decide whether to have %q
|
||
|
||
// characters
|
||
["%c", "x".charCodeAt(0), "x"],
|
||
["%c", 0xe4, "ä"],
|
||
["%c", 0x672c, "本"],
|
||
["%c", "日".charCodeAt(0), "日"],
|
||
// Specifying precision should have no effect.
|
||
["%.0c", "⌘".charCodeAt(0), "⌘"],
|
||
["%3c", "⌘".charCodeAt(0), " ⌘"],
|
||
["%-3c", "⌘".charCodeAt(0), "⌘ "],
|
||
|
||
// Runes that are not printable.
|
||
// {"%c", '\U00000e00', "\u0e00"}, // TODO check if \U escape exists in js
|
||
//["%c", '\U0010ffff'.codePointAt(0), "\U0010ffff"],
|
||
|
||
// Runes that are not valid.
|
||
["%c", -1, "<22>"],
|
||
// TODO surrogate half, doesn't make sense in itself, how
|
||
// to determine in JS?
|
||
// ["%c", 0xDC80, "<22>"],
|
||
["%c", 0x110000, "<22>"],
|
||
["%c", 0xfffffffff, "<22>"],
|
||
|
||
// TODO
|
||
// escaped characters
|
||
// Runes that are not printable.
|
||
// Runes that are not valid.
|
||
|
||
// width
|
||
["%5s", "abc", " abc"],
|
||
["%2s", "\u263a", " ☺"],
|
||
["%-5s", "abc", "abc "],
|
||
["%05s", "abc", "00abc"],
|
||
["%5s", "abcdefghijklmnopqrstuvwxyz", "abcdefghijklmnopqrstuvwxyz"],
|
||
["%.5s", "abcdefghijklmnopqrstuvwxyz", "abcde"],
|
||
["%.0s", "日本語日本語", ""],
|
||
["%.5s", "日本語日本語", "日本語日本"],
|
||
["%.10s", "日本語日本語", "日本語日本語"],
|
||
// ["%08q", "abc", `000"abc"`], // TODO verb q
|
||
// ["%-8q", "abc", `"abc" `],
|
||
//["%.5q", "abcdefghijklmnopqrstuvwxyz", `"abcde"`],
|
||
["%.5x", "abcdefghijklmnopqrstuvwxyz", "6162636465"],
|
||
//["%.3q", "日本語日本語", `"日本語"`],
|
||
//["%.1q", "日本語", `"日"`]
|
||
// change of go testcase utf-8([日]) = 0xe697a5, utf-16= 65e5 and
|
||
// our %x takes lower byte of string "%.1x", "日本語", "e6"],,
|
||
["%.1x", "日本語", "e5"],
|
||
//["%10.1q", "日本語日本語", ` "日"`],
|
||
// ["%10v", null, " <nil>"], // TODO null, undefined ...
|
||
// ["%-10v", null, "<nil> "],
|
||
|
||
// integers
|
||
["%d", 12345, "12345"],
|
||
["%d", -12345, "-12345"],
|
||
// ["%d", ^uint8(0), "255"],
|
||
//["%d", ^uint16(0), "65535"],
|
||
//["%d", ^uint32(0), "4294967295"],
|
||
//["%d", ^uint64(0), "18446744073709551615"],
|
||
["%d", -1 << 7, "-128"],
|
||
["%d", -1 << 15, "-32768"],
|
||
["%d", -1 << 31, "-2147483648"],
|
||
//["%d", (-1 << 63), "-9223372036854775808"],
|
||
["%.d", 0, ""],
|
||
["%.0d", 0, ""],
|
||
["%6.0d", 0, " "],
|
||
["%06.0d", 0, " "], // 0 flag should be ignored
|
||
["% d", 12345, " 12345"],
|
||
["%+d", 12345, "+12345"],
|
||
["%+d", -12345, "-12345"],
|
||
["%b", 7, "111"],
|
||
["%b", -6, "-110"],
|
||
// ["%b", ^uint32(0), "11111111111111111111111111111111"],
|
||
// ["%b", ^uint64(0),
|
||
// "1111111111111111111111111111111111111111111111111111111111111111"],
|
||
// ["%b", int64(-1 << 63), zeroFill("-1", 63, "")],
|
||
// 0 octal notation not allowed in struct node...
|
||
["%o", parseInt("01234", 8), "1234"],
|
||
["%#o", parseInt("01234", 8), "01234"],
|
||
// ["%o", ^uint32(0), "37777777777"],
|
||
// ["%o", ^uint64(0), "1777777777777777777777"],
|
||
["%#X", 0, "0X0"],
|
||
["%x", 0x12abcdef, "12abcdef"],
|
||
["%X", 0x12abcdef, "12ABCDEF"],
|
||
// ["%x", ^uint32(0), "ffffffff"],
|
||
// ["%X", ^uint64(0), "FFFFFFFFFFFFFFFF"],
|
||
["%.20b", 7, "00000000000000000111"],
|
||
["%10d", 12345, " 12345"],
|
||
["%10d", -12345, " -12345"],
|
||
["%+10d", 12345, " +12345"],
|
||
["%010d", 12345, "0000012345"],
|
||
["%010d", -12345, "-000012345"],
|
||
["%20.8d", 1234, " 00001234"],
|
||
["%20.8d", -1234, " -00001234"],
|
||
["%020.8d", 1234, " 00001234"],
|
||
["%020.8d", -1234, " -00001234"],
|
||
["%-20.8d", 1234, "00001234 "],
|
||
["%-20.8d", -1234, "-00001234 "],
|
||
["%-#20.8x", 0x1234abc, "0x01234abc "],
|
||
["%-#20.8X", 0x1234abc, "0X01234ABC "],
|
||
["%-#20.8o", parseInt("01234", 8), "00001234 "],
|
||
|
||
// Test correct f.intbuf overflow checks. // TODO, lazy
|
||
// unicode format // TODO, decide whether unicode verb makes sense %U
|
||
|
||
// floats
|
||
["%+.3e", 0.0, "+0.000e+00"],
|
||
["%+.3e", 1.0, "+1.000e+00"],
|
||
["%+.3f", -1.0, "-1.000"],
|
||
["%+.3F", -1.0, "-1.000"],
|
||
//["%+.3F", float32(-1.0), "-1.000"],
|
||
["%+07.2f", 1.0, "+001.00"],
|
||
["%+07.2f", -1.0, "-001.00"],
|
||
["%-07.2f", 1.0, "1.00 "],
|
||
["%-07.2f", -1.0, "-1.00 "],
|
||
["%+-07.2f", 1.0, "+1.00 "],
|
||
["%+-07.2f", -1.0, "-1.00 "],
|
||
["%-+07.2f", 1.0, "+1.00 "],
|
||
["%-+07.2f", -1.0, "-1.00 "],
|
||
["%+10.2f", +1.0, " +1.00"],
|
||
["%+10.2f", -1.0, " -1.00"],
|
||
["% .3E", -1.0, "-1.000E+00"],
|
||
["% .3e", 1.0, " 1.000e+00"],
|
||
["%+.3g", 0.0, "+0"],
|
||
["%+.3g", 1.0, "+1"],
|
||
["%+.3g", -1.0, "-1"],
|
||
["% .3g", -1.0, "-1"],
|
||
["% .3g", 1.0, " 1"],
|
||
// //["%b", float32(1.0), "8388608p-23"],
|
||
// ["%b", 1.0, "4503599627370496p-52"],
|
||
// // Test sharp flag used with floats.
|
||
["%#g", 1e-323, "1.00000e-323"],
|
||
["%#g", -1.0, "-1.00000"],
|
||
["%#g", 1.1, "1.10000"],
|
||
["%#g", 123456.0, "123456."],
|
||
//["%#g", 1234567.0, "1.234567e+06"],
|
||
// the line above is incorrect in go (according to
|
||
// my posix reading) %f-> prec = prec-1
|
||
["%#g", 1234567.0, "1.23457e+06"],
|
||
["%#g", 1230000.0, "1.23000e+06"],
|
||
["%#g", 1000000.0, "1.00000e+06"],
|
||
["%#.0f", 1.0, "1."],
|
||
["%#.0e", 1.0, "1.e+00"],
|
||
["%#.0g", 1.0, "1."],
|
||
["%#.0g", 1100000.0, "1.e+06"],
|
||
["%#.4f", 1.0, "1.0000"],
|
||
["%#.4e", 1.0, "1.0000e+00"],
|
||
["%#.4g", 1.0, "1.000"],
|
||
["%#.4g", 100000.0, "1.000e+05"],
|
||
["%#.0f", 123.0, "123."],
|
||
["%#.0e", 123.0, "1.e+02"],
|
||
["%#.0g", 123.0, "1.e+02"],
|
||
["%#.4f", 123.0, "123.0000"],
|
||
["%#.4e", 123.0, "1.2300e+02"],
|
||
["%#.4g", 123.0, "123.0"],
|
||
["%#.4g", 123000.0, "1.230e+05"],
|
||
["%#9.4g", 1.0, " 1.000"],
|
||
// The sharp flag has no effect for binary float format.
|
||
// ["%#b", 1.0, "4503599627370496p-52"], // TODO binary for floats
|
||
// Precision has no effect for binary float format.
|
||
//["%.4b", float32(1.0), "8388608p-23"], // TODO s.above
|
||
// ["%.4b", -1.0, "-4503599627370496p-52"],
|
||
// Test correct f.intbuf boundary checks.
|
||
//["%.68f", 1.0, zeroFill("1.", 68, "")], // TODO zerofill
|
||
//["%.68f", -1.0, zeroFill("-1.", 68, "")], //TODO s.a.
|
||
// float infinites and NaNs
|
||
["%f", Number.POSITIVE_INFINITY, "+Inf"],
|
||
["%.1f", Number.NEGATIVE_INFINITY, "-Inf"],
|
||
["% f", NaN, " NaN"],
|
||
["%20f", Number.POSITIVE_INFINITY, " +Inf"],
|
||
// ["% 20F", Number.POSITIVE_INFINITY, " Inf"], // TODO : wut?
|
||
["% 20e", Number.NEGATIVE_INFINITY, " -Inf"],
|
||
["%+20E", Number.NEGATIVE_INFINITY, " -Inf"],
|
||
["% +20g", Number.NEGATIVE_INFINITY, " -Inf"],
|
||
["%+-20G", Number.POSITIVE_INFINITY, "+Inf "],
|
||
["%20e", NaN, " NaN"],
|
||
["% +20E", NaN, " +NaN"],
|
||
["% -20g", NaN, " NaN "],
|
||
["%+-20G", NaN, "+NaN "],
|
||
// Zero padding does not apply to infinities and NaN.
|
||
["%+020e", Number.POSITIVE_INFINITY, " +Inf"],
|
||
["%-020f", Number.NEGATIVE_INFINITY, "-Inf "],
|
||
["%-020E", NaN, "NaN "],
|
||
|
||
// complex values // go specific
|
||
// old test/fmt_test.go
|
||
["%e", 1.0, "1.000000e+00"],
|
||
["%e", 1234.5678e3, "1.234568e+06"],
|
||
["%e", 1234.5678e-8, "1.234568e-05"],
|
||
["%e", -7.0, "-7.000000e+00"],
|
||
["%e", -1e-9, "-1.000000e-09"],
|
||
["%f", 1234.5678e3, "1234567.800000"],
|
||
["%f", 1234.5678e-8, "0.000012"],
|
||
["%f", -7.0, "-7.000000"],
|
||
["%f", -1e-9, "-0.000000"],
|
||
// ["%g", 1234.5678e3, "1.2345678e+06"],
|
||
// I believe the above test from go is incorrect according to posix, s. above.
|
||
["%g", 1234.5678e3, "1.23457e+06"],
|
||
//["%g", float32(1234.5678e3), "1.2345678e+06"],
|
||
//["%g", 1234.5678e-8, "1.2345678e-05"], // posix, see above
|
||
["%g", 1234.5678e-8, "1.23457e-05"],
|
||
["%g", -7.0, "-7"],
|
||
["%g", -1e-9, "-1e-09"],
|
||
//["%g", float32(-1e-9), "-1e-09"],
|
||
["%E", 1.0, "1.000000E+00"],
|
||
["%E", 1234.5678e3, "1.234568E+06"],
|
||
["%E", 1234.5678e-8, "1.234568E-05"],
|
||
["%E", -7.0, "-7.000000E+00"],
|
||
["%E", -1e-9, "-1.000000E-09"],
|
||
//["%G", 1234.5678e3, "1.2345678E+06"], // posix, see above
|
||
["%G", 1234.5678e3, "1.23457E+06"],
|
||
//["%G", float32(1234.5678e3), "1.2345678E+06"],
|
||
//["%G", 1234.5678e-8, "1.2345678E-05"], // posic, see above
|
||
["%G", 1234.5678e-8, "1.23457E-05"],
|
||
["%G", -7.0, "-7"],
|
||
["%G", -1e-9, "-1E-09"],
|
||
//["%G", float32(-1e-9), "-1E-09"],
|
||
["%20.5s", "qwertyuiop", " qwert"],
|
||
["%.5s", "qwertyuiop", "qwert"],
|
||
["%-20.5s", "qwertyuiop", "qwert "],
|
||
["%20c", "x".charCodeAt(0), " x"],
|
||
["%-20c", "x".charCodeAt(0), "x "],
|
||
["%20.6e", 1.2345e3, " 1.234500e+03"],
|
||
["%20.6e", 1.2345e-3, " 1.234500e-03"],
|
||
["%20e", 1.2345e3, " 1.234500e+03"],
|
||
["%20e", 1.2345e-3, " 1.234500e-03"],
|
||
["%20.8e", 1.2345e3, " 1.23450000e+03"],
|
||
["%20f", 1.23456789e3, " 1234.567890"],
|
||
["%20f", 1.23456789e-3, " 0.001235"],
|
||
["%20f", 12345678901.23456789, " 12345678901.234568"],
|
||
["%-20f", 1.23456789e3, "1234.567890 "],
|
||
["%20.8f", 1.23456789e3, " 1234.56789000"],
|
||
["%20.8f", 1.23456789e-3, " 0.00123457"],
|
||
// ["%g", 1.23456789e3, "1234.56789"],
|
||
// posix ... precision(2) = precision(def=6) - (exp(3)+1)
|
||
["%g", 1.23456789e3, "1234.57"],
|
||
// ["%g", 1.23456789e-3, "0.00123456789"], posix...
|
||
["%g", 1.23456789e-3, "0.00123457"], // see above prec6 = precdef6 - (-3+1)
|
||
//["%g", 1.23456789e20, "1.23456789e+20"],
|
||
["%g", 1.23456789e20, "1.23457e+20"],
|
||
|
||
// arrays // TODO
|
||
// slice : go specific
|
||
|
||
// TODO decide how to handle deeper types, arrays, objects
|
||
// byte arrays and slices with %b,%c,%d,%o,%U and %v
|
||
// f.space should and f.plus should not have an effect with %v.
|
||
// f.space and f.plus should have an effect with %d.
|
||
|
||
// Padding with byte slices.
|
||
// Same for strings
|
||
["%2x", "", " "], // 103
|
||
["%#2x", "", " "],
|
||
["% 02x", "", "00"],
|
||
["%# 02x", "", "00"],
|
||
["%-2x", "", " "],
|
||
["%-02x", "", " "],
|
||
["%8x", "\xab", " ab"],
|
||
["% 8x", "\xab", " ab"],
|
||
["%#8x", "\xab", " 0xab"],
|
||
["%# 8x", "\xab", " 0xab"],
|
||
["%08x", "\xab", "000000ab"],
|
||
["% 08x", "\xab", "000000ab"],
|
||
["%#08x", "\xab", "00000xab"],
|
||
["%# 08x", "\xab", "00000xab"],
|
||
["%10x", "\xab\xcd", " abcd"],
|
||
["% 10x", "\xab\xcd", " ab cd"],
|
||
["%#10x", "\xab\xcd", " 0xabcd"],
|
||
["%# 10x", "\xab\xcd", " 0xab 0xcd"],
|
||
["%010x", "\xab\xcd", "000000abcd"],
|
||
["% 010x", "\xab\xcd", "00000ab cd"],
|
||
["%#010x", "\xab\xcd", "00000xabcd"],
|
||
["%# 010x", "\xab\xcd", "00xab 0xcd"],
|
||
["%-10X", "\xab", "AB "],
|
||
["% -010X", "\xab", "AB "],
|
||
["%#-10X", "\xab\xcd", "0XABCD "],
|
||
["%# -010X", "\xab\xcd", "0XAB 0XCD "],
|
||
|
||
// renamings
|
||
// Formatter
|
||
// GoStringer
|
||
|
||
// %T TODO possibly %#T object(constructor)
|
||
["%T", {}, "object"],
|
||
["%T", 1, "number"],
|
||
["%T", "", "string"],
|
||
["%T", undefined, "undefined"],
|
||
["%T", null, "object"],
|
||
["%T", S, "function"],
|
||
["%T", true, "boolean"],
|
||
["%T", Symbol(), "symbol"],
|
||
|
||
// %p with pointers
|
||
|
||
// erroneous things
|
||
// {"", nil, "%!(EXTRA <nil>)"},
|
||
// {"", 2, "%!(EXTRA int=2)"},
|
||
// {"no args", "hello", "no args%!(EXTRA string=hello)"},
|
||
// {"%s %", "hello", "hello %!(NOVERB)"},
|
||
// {"%s %.2", "hello", "hello %!(NOVERB)"},
|
||
// {"%017091901790959340919092959340919017929593813360", 0,
|
||
// "%!(NOVERB)%!(EXTRA int=0)"},
|
||
// {"%184467440737095516170v", 0, "%!(NOVERB)%!(EXTRA int=0)"},
|
||
// // Extra argument errors should format without flags set.
|
||
// {"%010.2", "12345", "%!(NOVERB)%!(EXTRA string=12345)"},
|
||
//
|
||
// // Test that maps with non-reflexive keys print all keys and values.
|
||
// {"%v", map[float64]int{NaN: 1, NaN: 1}, "map[NaN:1 NaN:1]"},
|
||
|
||
// more floats
|
||
|
||
["%.2f", 1.0, "1.00"],
|
||
["%.2f", -1.0, "-1.00"],
|
||
["% .2f", 1.0, " 1.00"],
|
||
["% .2f", -1.0, "-1.00"],
|
||
["%+.2f", 1.0, "+1.00"],
|
||
["%+.2f", -1.0, "-1.00"],
|
||
["%7.2f", 1.0, " 1.00"],
|
||
["%7.2f", -1.0, " -1.00"],
|
||
["% 7.2f", 1.0, " 1.00"],
|
||
["% 7.2f", -1.0, " -1.00"],
|
||
["%+7.2f", 1.0, " +1.00"],
|
||
["%+7.2f", -1.0, " -1.00"],
|
||
["% +7.2f", 1.0, " +1.00"],
|
||
["% +7.2f", -1.0, " -1.00"],
|
||
["%07.2f", 1.0, "0001.00"],
|
||
["%07.2f", -1.0, "-001.00"],
|
||
["% 07.2f", 1.0, " 001.00"], //153 here
|
||
["% 07.2f", -1.0, "-001.00"],
|
||
["%+07.2f", 1.0, "+001.00"],
|
||
["%+07.2f", -1.0, "-001.00"],
|
||
["% +07.2f", 1.0, "+001.00"],
|
||
["% +07.2f", -1.0, "-001.00"]
|
||
];
|
||
|
||
test(function testThorough(): void {
|
||
tests.forEach((t, i): void => {
|
||
// p(t)
|
||
const is = S(t[0], t[1]);
|
||
const should = t[2];
|
||
assertEquals(
|
||
is,
|
||
should,
|
||
`failed case[${i}] : is >${is}< should >${should}<`
|
||
);
|
||
});
|
||
});
|
||
|
||
test(function testWeirdos(): void {
|
||
assertEquals(S("%.d", 9), "9");
|
||
assertEquals(
|
||
S("dec[%d]=%d hex[%[1]d]=%#x oct[%[1]d]=%#o %s", 1, 255, "Third"),
|
||
"dec[1]=255 hex[1]=0xff oct[1]=0377 Third"
|
||
);
|
||
});
|
||
|
||
test(function formatV(): void {
|
||
const a = { a: { a: { a: { a: { a: { a: { a: {} } } } } } } };
|
||
assertEquals(S("%v", a), "[object Object]");
|
||
assertEquals(S("%#v", a), "{ a: { a: { a: { a: [Object] } } } }");
|
||
assertEquals(
|
||
S("%#.8v", a),
|
||
"{ a: { a: { a: { a: { a: { a: { a: {} } } } } } } }"
|
||
);
|
||
assertEquals(S("%#.1v", a), "{ a: [Object] }");
|
||
});
|
||
|
||
test(function formatJ(): void {
|
||
const a = { a: { a: { a: { a: { a: { a: { a: {} } } } } } } };
|
||
assertEquals(S("%j", a), `{"a":{"a":{"a":{"a":{"a":{"a":{"a":{}}}}}}}}`);
|
||
});
|
||
|
||
test(function flagLessThan(): void {
|
||
const a = { a: { a: { a: { a: { a: { a: { a: {} } } } } } } };
|
||
const aArray = [a, a, a];
|
||
assertEquals(
|
||
S("%<#.1v", aArray),
|
||
"[ { a: [Object] }, { a: [Object] }, { a: [Object] } ]"
|
||
);
|
||
const fArray = [1.2345, 0.98765, 123456789.5678];
|
||
assertEquals(S("%<.2f", fArray), "[ 1.23, 0.99, 123456789.57 ]");
|
||
});
|
||
|
||
test(function testErrors(): void {
|
||
// wrong type : TODO strict mode ...
|
||
//assertEquals(S("%f", "not a number"), "%!(BADTYPE flag=f type=string)")
|
||
assertEquals(S("A %h", ""), "A %!(BAD VERB 'h')");
|
||
assertEquals(S("%J", ""), "%!(BAD VERB 'J')");
|
||
assertEquals(S("bla%J", ""), "bla%!(BAD VERB 'J')");
|
||
assertEquals(S("%Jbla", ""), "%!(BAD VERB 'J')bla");
|
||
|
||
assertEquals(S("%d"), "%!(MISSING 'd')");
|
||
assertEquals(S("%d %d", 1), "1 %!(MISSING 'd')");
|
||
assertEquals(S("%d %f A", 1), "1 %!(MISSING 'f') A");
|
||
|
||
assertEquals(S("%*.2f", "a", 1.1), "%!(BAD WIDTH 'a')");
|
||
assertEquals(S("%.*f", "a", 1.1), "%!(BAD PREC 'a')");
|
||
assertEquals(S("%.[2]*f", 1.23, "p"), "%!(BAD PREC 'p')%!(EXTRA '1.23')");
|
||
assertEquals(S("%.[2]*[1]f Yippie!", 1.23, "p"), "%!(BAD PREC 'p') Yippie!");
|
||
|
||
assertEquals(S("%[1]*.2f", "a", "p"), "%!(BAD WIDTH 'a')");
|
||
|
||
assertEquals(S("A", "a", "p"), "A%!(EXTRA 'a' 'p')");
|
||
assertEquals(S("%[2]s %[2]s", "a", "p"), "p p%!(EXTRA 'a')");
|
||
|
||
// remains to be determined how to handle bad indices ...
|
||
// (realistically) the entire error handling is still up for grabs.
|
||
assertEquals(S("%[hallo]s %d %d %d", 1, 2, 3, 4), "%!(BAD INDEX) 2 3 4");
|
||
assertEquals(S("%[5]s", 1, 2, 3, 4), "%!(BAD INDEX)%!(EXTRA '2' '3' '4')");
|
||
assertEquals(S("%[5]f"), "%!(BAD INDEX)");
|
||
assertEquals(S("%.[5]f"), "%!(BAD INDEX)");
|
||
assertEquals(S("%.[5]*f"), "%!(BAD INDEX)");
|
||
});
|
||
|
||
runIfMain(import.meta);
|