1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-27 04:17:35 +00:00

AK: Borrow exact format syntax form std::format.

Instead of just implementing format specifiers ad-hog this commit
implements the exact syntax std::format uses.

There are still a ton of features that are not supported by this
implementation, however, the format specifiers should be parsed
correctly.

In some cases however, the format specifiers aren't quite parsed
correctly, for example:

    String::formatted("{:{}}", 42, 4)

should produce the string "  42" however an (unrelated) assertion fails.
This is because vformat doesn't consider nested parentheses. I have to
spend some time coming up with a simple way of doing this, I don't feel
like doing that right now.

The fundamental code for this already exists, by limiting the number of
format arguments (arbitrarily) to 256 large widths are used to encode
that these should be taken from other format parameters.
This commit is contained in:
asynts 2020-09-25 13:11:29 +02:00 committed by Andreas Kling
parent 84d276dba0
commit 2030084746
2 changed files with 203 additions and 60 deletions

View file

@ -73,28 +73,32 @@ static void write_escaped_literal(StringBuilder& builder, StringView literal)
}
}
static size_t parse_number(StringView input)
static bool parse_number(GenericLexer& lexer, size_t& value)
{
size_t value = 0;
value = 0;
for (char ch : input) {
value *= 10;
value += ch - '0';
bool consumed_at_least_one = false;
while (!lexer.is_eof()) {
if (lexer.next_is(is_digit)) {
value *= 10;
value += lexer.consume() - '0';
consumed_at_least_one = true;
} else {
break;
}
}
return value;
return consumed_at_least_one;
}
constexpr size_t use_next_index = NumericLimits<size_t>::max();
static bool parse_format_specifier(StringView input, FormatSpecifier& specifier)
{
specifier.index = NumericLimits<size_t>::max();
GenericLexer lexer { input };
auto index = lexer.consume_while([](char ch) { return StringView { "0123456789" }.contains(ch); });
if (index.length() > 0)
specifier.index = parse_number(index);
if (!parse_number(lexer, specifier.index))
specifier.index = use_next_index;
if (!lexer.consume_specific(':'))
return lexer.is_eof();
@ -103,6 +107,20 @@ static bool parse_format_specifier(StringView input, FormatSpecifier& specifier)
return true;
}
static bool parse_nested_replacement_field(GenericLexer& lexer, size_t& index)
{
if (!lexer.consume_specific('{'))
return false;
if (!parse_number(lexer, index))
index = use_next_index;
if (!lexer.consume_specific('}'))
ASSERT_NOT_REACHED();
return true;
}
} // namespace
namespace AK {
@ -137,8 +155,7 @@ void vformat(StringBuilder& builder, StringView fmtstr, AK::Span<const TypeErase
ASSERT_NOT_REACHED();
auto& parameter = parameters[specifier.index];
if (!parameter.formatter(builder, parameter.value, specifier.flags))
ASSERT_NOT_REACHED();
parameter.formatter(builder, parameter.value, specifier.flags, parameters);
vformat(builder, fmtstr.substring_view(closing + 1), parameters, argument_index);
}
@ -149,43 +166,133 @@ void vformat(const LogStream& stream, StringView fmtstr, Span<const TypeErasedPa
stream << builder.to_string();
}
bool Formatter<StringView>::parse(StringView flags)
void StandardFormatter::parse(StringView specifier)
{
return flags.is_empty();
GenericLexer lexer { specifier };
if (StringView { "<^>" }.contains(lexer.peek(1))) {
ASSERT(!lexer.next_is(is_any_of("{}")));
m_fill = lexer.consume();
}
if (lexer.consume_specific('<'))
m_align = Align::Left;
else if (lexer.consume_specific('^'))
m_align = Align::Center;
else if (lexer.consume_specific('>'))
m_align = Align::Right;
if (lexer.consume_specific('-'))
m_sign = Sign::NegativeOnly;
else if (lexer.consume_specific('+'))
m_sign = Sign::PositiveAndNegative;
else if (lexer.consume_specific(' '))
m_sign = Sign::ReserveSpace;
if (lexer.consume_specific('#'))
m_alternative_form = true;
if (lexer.consume_specific('0'))
m_zero_pad = true;
if (size_t index = 0; parse_nested_replacement_field(lexer, index))
m_width = value_from_arg + index;
else if (size_t width = 0; parse_number(lexer, width))
m_width = width;
if (lexer.consume_specific('.')) {
if (size_t index = 0; parse_nested_replacement_field(lexer, index))
m_precision = value_from_arg + index;
else if (size_t precision = 0; parse_number(lexer, precision))
m_precision = precision;
}
if (lexer.consume_specific('b'))
m_mode = Mode::Binary;
else if (lexer.consume_specific('d'))
m_mode = Mode::Decimal;
else if (lexer.consume_specific('o'))
m_mode = Mode::Octal;
else if (lexer.consume_specific('x'))
m_mode = Mode::Hexadecimal;
else if (lexer.consume_specific('c'))
m_mode = Mode::Character;
else if (lexer.consume_specific('s'))
m_mode = Mode::String;
else if (lexer.consume_specific('p'))
m_mode = Mode::Pointer;
if (!lexer.is_eof())
dbg() << __PRETTY_FUNCTION__ << " did not consume '" << lexer.remaining() << "'";
ASSERT(lexer.is_eof());
}
void Formatter<StringView>::format(StringBuilder& builder, StringView value)
void Formatter<StringView>::format(StringBuilder& builder, StringView value, Span<const TypeErasedParameter>)
{
if (m_align != Align::Default)
TODO();
if (m_sign != Sign::Default)
ASSERT_NOT_REACHED();
if (m_alternative_form)
ASSERT_NOT_REACHED();
if (m_zero_pad)
ASSERT_NOT_REACHED();
if (m_width != value_not_set)
TODO();
if (m_precision != value_not_set)
TODO();
if (m_mode != Mode::Default && m_mode != Mode::String)
ASSERT_NOT_REACHED();
builder.append(value);
}
template<typename T>
bool Formatter<T, typename EnableIf<IsIntegral<T>::value>::Type>::parse(StringView flags)
void Formatter<T, typename EnableIf<IsIntegral<T>::value>::Type>::format(StringBuilder& builder, T value, Span<const TypeErasedParameter> parameters)
{
GenericLexer lexer { flags };
if (m_align != Align::Default)
TODO();
if (m_sign != Sign::Default)
TODO();
if (m_alternative_form)
TODO();
if (m_precision != value_not_set)
ASSERT_NOT_REACHED();
if (m_mode == Mode::Character)
TODO();
if (lexer.consume_specific('0'))
zero_pad = true;
auto field_width = lexer.consume_while([](char ch) { return StringView { "0123456789" }.contains(ch); });
if (field_width.length() > 0)
this->field_width = parse_number(field_width);
if (lexer.consume_specific('x'))
hexadecimal = true;
return lexer.is_eof();
}
template<typename T>
void Formatter<T, typename EnableIf<IsIntegral<T>::value>::Type>::format(StringBuilder& builder, T value)
{
char* bufptr;
if (hexadecimal)
PrintfImplementation::print_hex([&](auto, char ch) { builder.append(ch); }, bufptr, value, false, false, false, zero_pad, field_width);
else if (IsSame<typename MakeUnsigned<T>::Type, T>::value)
PrintfImplementation::print_u64([&](auto, char ch) { builder.append(ch); }, bufptr, value, false, zero_pad, field_width);
int base;
if (m_mode == Mode::Binary)
TODO();
else if (m_mode == Mode::Octal)
TODO();
else if (m_mode == Mode::Decimal || m_mode == Mode::Default)
base = 10;
else if (m_mode == Mode::Hexadecimal)
base = 16;
else
PrintfImplementation::print_i64([&](auto, char ch) { builder.append(ch); }, bufptr, value, false, zero_pad, field_width);
ASSERT_NOT_REACHED();
size_t width = m_width;
if (m_width >= value_from_arg) {
const auto parameter = parameters.at(m_width - value_from_arg);
// FIXME: Totally unsave cast. We should store the type in TypeErasedParameter. For compactness it could be smart to
// find a few addresses that can not be valid function pointers and encode the type information there?
width = *reinterpret_cast<const size_t*>(parameter.value);
}
// FIXME: We really need one canonical print implementation that just takes a base.
(void)base;
char* bufptr;
if (m_mode == Mode::Hexadecimal)
PrintfImplementation::print_hex([&](auto, char ch) { builder.append(ch); }, bufptr, value, false, false, false, m_zero_pad, width);
else if (IsSame<typename MakeUnsigned<T>::Type, T>::value)
PrintfImplementation::print_u64([&](auto, char ch) { builder.append(ch); }, bufptr, value, false, m_zero_pad, width);
else
PrintfImplementation::print_i64([&](auto, char ch) { builder.append(ch); }, bufptr, value, false, m_zero_pad, width);
}
template struct Formatter<unsigned char, void>;