An editable document only holds valid UTF-8, but copy_scalar() copied the
strings and keys of a view of another document unchecked. A document of
a weaker check (or a borrowed text that changed after parsing) could
therefore bring ill-formed UTF-8 into it. The copy is checked now, with
the error that dump() reports for the string; copies within the same
document stay unchecked.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Counting and copying the nodes of a view or a basic_json value into an
editable document recursed once per nesting level, so that a deeply
nested value overflowed the stack. The four functions now walk the value
with an explicit stack, as materialize() does.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
basic_json_document gets a second template parameter, Editable
(false by default), plus the aliases json_editable_document,
json_editable_view, ordered_json_editable_document and
ordered_json_editable_view.
Editable documents can change values and structure without
rewriting the source text: set()/push_back() on values, keys,
array indices and JSON pointers; insert() before an array
element; erase() of an object key, array index or JSON pointer.
New values and element sequences go into edit storage that the
document owns and never moves, so views keep referring to their
value across edits and a parsed node never moves. Read-only
documents walk the plain node array and are unaffected.
Strings are checked for UTF-8 on entry, so dump() of an editable
document never throws type_error.316. Binary values cannot be
stored (type_error.319).
A seeded differential test applies random edits to an editable
document and to the equivalent ordered_json and compares both
after every step.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The dump and comparison tests moved to their own file in the dump pull
request; the hash index tests of this branch join them there.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Lookups in objects with 128 members or more return the first member of a
repeated key again, like the linear search of smaller objects. This undoes
the code change of a69542046; its test now expects the first member from
lookups (with and without a table) and the last value from materialize()
and basic_json::parse().
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Lookups in objects with 128 members or more now return the last member of
a repeated key, like the linear search of smaller objects and like
materialize() and basic_json::parse(). build_object_index let the first
occurrence win, so the same text gave different results depending on the
size of the object.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
shrink_to_fit() now trims the tables of large objects like the node array and the decoded strings, and the list of large objects is released as soon as the tables are built.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The key hash is not seeded, so keys chosen to collide made building the table quadratic (20,000 colliding keys took 470 ms to parse). A key may now sit at most 64 slots from its home slot; if a key would sit further away, the table is dropped and the object is searched linearly. Lookups stop after the same distance.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Speed up json_view's parser with SIMD scanning and a hash table
for large objects.
Long runs of string bytes are scanned 16 bytes at a time with NEON
(AArch64, GCC and Clang) and SSE2 (x86-64), both baseline
instruction sets. Keys keep 16 table checks before the vector
loop, because their lengths repeat from record to record; string
values get 8, because their lengths vary more. Non-ASCII text is
validated 16 bytes at a time with simdjson's "lookup4" check
(Keiser and Lemire, 2021), with NEON on AArch64 and, on x86-64,
with SSSE3. SSSE3 is not part of baseline x86-64, so the check is
compiled for SSSE3 with a function attribute and used only where
CPUID reports it, which all x86-64 CPUs since about 2011 do; the
answer is cached in a statically initialized atomic, so there is
no guard of a local static and no global constructor. The same
input is accepted either way. JSON_VIEW_NO_SIMD selects the
portable code.
On x86-64, string runs are now checked vector-first: one SSE2
compare from the first byte finds the end of most keys and short
values, instead of a branch per byte for the first 8-16 bytes.
AArch64 keeps the byte-wise steps, where a NEON mask costs more and
the branches predict well. Entering an object or array no longer
stalls: open() stores the parent's frame field by field instead of
building it on the stack and reading it back with wider loads,
which waited for the narrower stores to retire.
Objects with 128 members or more get an open-addressing hash table
built when the object closes, so operator[], at(), find(),
contains(), count(), value(), and JSON pointers take constant time
on average in such objects; of duplicate keys, the first is kept,
as for the linear search. The idea comes from Boost.JSON.
simdjson is credited in simd.hpp's SPDX block, the README, and
license.md.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The MinGW linker of the Windows clang jobs cannot link object files with more
than 32767 sections ("relocation truncated to fit: IMAGE_REL_AMD64_REL32
against `.rdata'"). unit-json_view.cpp reaches that limit as the stack
grows, so its "json_view dump" and "json_view comparison" test cases move
into unit-json_view_dump.cpp. The test generator and has_duplicate_keys()
that both files use move into json_view_test_helpers.hpp.
The new file mentions JSON_HAS_CPP_17, so it is built for C++17 like the file
it was split from.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
dump() writes every member and == resolves duplicate keys as parse()
does, whereas lookups find the first member of a duplicate key.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The source extent of a value was read from the next node, falling back to the rest of the document when that node held a decoded string. dump() of a small value could thus allocate a buffer as large as the document. Skip a few such nodes, cap the fallback estimate, and shrink a buffer that is much larger than its output.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Add basic_json_view::dump() and the comparison operators, and read
floats from the parser's digit layout instead of rescanning the
token.
dump(indent, indent_char, ensure_ascii, number_format) writes a
value the way ordered_json::parse(text).dump() writes it for the
same arguments: members in document order, all of them should a
key occur more than once; strings escaped by the same rules, using
the library's scanning kernels; floats written with the library's
to_chars conversion, so the output equals basic_json's byte for
byte; integers copied from the source, where they are already
canonical, except -0, which parse() reads as 0. There is no
error_handler argument, because the view only holds valid UTF-8.
number_format::source copies numbers exactly as they appear in the
source (e.g. "1.50", "1E2", "-0"), which basic_json cannot provide.
operator<< takes the indentation from the stream width, as for
basic_json. The writer walks iteratively, so nesting depth is
limited by memory only.
operator== and operator!= compare two views, or a view and a
basic_json value in either order, by the rules basic_json's
operator== uses: numbers compare by value across their types,
objects compare by their members with duplicate keys resolved as
parse() resolves them, member order matters only where the object
type keeps one, and discarded views compare as discarded basic_json
values do, including under JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON.
Nothing is materialized except single scalars.
While parsing, the view now records where the integer digits, the
fraction digits, and the exponent of a float token are, so floats
and doubles with at most 19 digits are read from that layout with
the library's decimal_to_float() instead of rescanning the token.
Both round correctly, so the values are those of parse(). get<double>(),
materialize(), dump(), and the comparisons all use it.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The same changes as on the dump branch, where they were first made, so
that this branch passes clang-tidy on its own.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Looking up the last member of a duplicate key cannot stop at a match, so
every lookup scanned the whole object (1.6 to 3.4 times slower for small
objects). operator[](key), at, find, value, contains, count, and JSON
pointer resolution return the first member again, as yyjson and simdjson
do; materialize() and get<map>() keep the last value, as parse().
The documentation says so in the feature page and on each lookup page,
and explains how to get the value parse() would give. The integer index
templates and the discarded chaining of operator[] stay.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
A view is tested with is_discarded(); the lookups in at(), value() and
contains(json_pointer) and the unit tests no longer rely on the explicit
conversion to bool.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Lookups (operator[], at, find, contains, count, value, JSON pointers)
return the last member of a duplicate key, as materialize() and
parse() keep it.
* operator[] on a discarded view returns a discarded view instead of
throwing, so v["a"]["b"] is safe for a missing "a".
* operator[] and at() take any integer type (not only int and size_t),
fixing ambiguous calls with unsigned, long, std::int64_t, ...
* Fix the operator[] documentation, which claimed a discarded view for
a type mismatch where type_error.305 is thrown.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Give basic_json_view the read-only access functions of basic_json:
operator[] and at() with keys and indices, front()/back(), find(),
contains(), count(), begin()/end() and cbegin()/cend(), items()
with structured bindings from C++17 on, and type_name().
Exceptions have the ids and messages of the const functions of
basic_json. Where basic_json has undefined behavior the view
answers safely: operator[] with a missing key or an out-of-range
index returns a discarded view, and front()/back() of an empty
container throw invalid_iterator.214. Objects are iterated in
document order, and all members are visited; duplicate-key lookups
find the first member (as yyjson and simdjson do), while parse(),
materialize(), and the map conversions keep the last value, as
parse() does. Keys of up to 16 bytes are compared with two
overlapping loads.
Add value conversions: get<T>()/get_to() for arithmetic types,
bool, nullptr_t, strings (std::basic_string copied,
string_view_t without a copy), BasicJsonType, views, std::vector,
and maps with string keys; get_string() for the string without a
copy; number_token() for the number exactly as written in the
source; value() with keys and JSON pointers; and operator[]/at()/
contains() with JSON pointers. Everything else, including types
with from_json(), goes through materialize() of that subtree.
get<T>() of arithmetic types is inlined down to the conversion, so
reading an integer needs no call.
detail::json_pointer_access exposes a pointer's reference tokens
to code outside basic_json.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Changes promised in the replies to the review of #5620:
- read() reuses the owned copy: a copied input (an rvalue byte container,
a stream, FILE*, wide characters, an iterator range) is appended to the
cleared owned buffer instead of a fresh std::string being move-assigned
over it. A moved-in std::string still replaces it.
- The error replay runs the library parser with a SAX handler that stores
nothing and throws the reported exception, instead of
BasicJsonType::parse(), which built a DOM up to the error.
- Pointers to and arrays of wide characters, and FILE*, take the input
adapter path (converted to UTF-8 into an owned buffer) instead of
failing a static_assert. Wide pointers are NUL-terminated.
- Streams are read in blocks through rdbuf() into the buffer, instead of
character by character through the input adapter; eofbit is set at the
end, as by the library's stream adapter.
- tape/tape_size/tape_cap are now nodes/node_count/node_capacity.
- The builder's members get descriptive, m_-prefixed names.
- The shallow frame stack is a std::array (in a wrapper with a
user-provided constructor: -Weffc++ wants the member initialized, and
value-initializing the 64 frames costs up to 30 % on small documents).
- The value dispatch macro has no NEXT parameter; the switch is left with
break, and the macro is undefined after its last use.
- literal() and digits_exceed() deduce the length from the string literal;
each literal has one helper; the two `if (!is_float)` blocks are merged.
- arena_ensure() sits right before the writes; the decode arena is a
std::vector<char>.
- The builder records the deepest nesting, and materialize() sizes its
stack from it once (growing it only if a subtree is deeper).
- Comment fixes (hex4, slow_string, the NUL rule, put_utf8, 0 - m, the
arena resize, grow()) and formatting (macro backslashes, digit cases,
one power of ten per line, value_t assertion next to is_container()).
- build_with's template parameters follow the builder's order.
Greg's idea of keeping the source offset in `off` and the arena offset in
`next` for escaped strings was measured and not adopted: string access got
7-18 % slower with a select, and 3.5-18 % slower with a branch-free field
pick (Apple M-series, clang -O3), parsing was unchanged.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
const_text is const, so std::move does not move from it; the second call
is part of the const rvalue test.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The fuzzer only parsed a std::string with default options. Also parse an
exact-size byte vector, which has no NUL after its last byte and takes the
bounds-checked path, and derive ignore_comments and ignore_trailing_commas
from the first input byte, comparing against json::parse and json::accept
with the same options.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The check rejects inputs of 0xFFFFFFF0 bytes or more, but the exception
message and the documentation said 4 GiB. Name the limit once
(max_input_size), and state 4 GiB minus 16 bytes in the message and the
documentation. Test the limit with a container that only claims the size.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
reserve(n) and the growth of the index computed n * sizeof(node) without a
check, which wraps around on 32-bit targets for inputs of about 1 GiB and
allocates a too small array. Throw std::bad_alloc for a count beyond the
address space and clamp the growth step to it.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
The non-little-endian path of the node writer wrote the integer's native
word over len and next, so len got the high half there. Compose and split
the value explicitly (len is the low half, next the high half); the
little-endian path stays a plain memcpy.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
auto v = json_document::parse(text).root() compiled and left the view
dangling. Delete the overload for rvalue documents, take a named document
in the tests, and document the lifetime rule.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
explicit operator bool meant "refers to a value", which silently differs
from what a basic_json converts to. Use !v.is_discarded() instead.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
json_document::parse(std::move(const_string)) failed to compile with
"no matching read_kind": only a non-const rvalue std::string can be moved
from. Treat a const rvalue as a copied byte container.
parse() does not accept everything BasicJsonType::parse() does: a FILE*
and pointers to or arrays of wide characters are rejected at compile time.
List the supported inputs instead.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
parse(ptr, len) compiled: len converted to allow_exceptions, and ptr was
read as a C string, past the end of a buffer without a terminating NUL.
Delete the overloads of parse, parse_copy, accept, and read that take an
integer other than bool where the flags are expected.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Add json_document and json_view, a read-only, zero-copy index of a
JSON text, as the first public slice of the zero-copy view (#5295).
A parse produces a flat array of 16-byte nodes in document order,
one per value and one per object key. Strings stay in the source
text; escaped strings are decoded into an arena. Integers are
converted while their digits are in the cache; floats keep only
their digit layout and are converted on read. Containers store the
size of their subtree, so a reader can step over one in constant
time. A document makes a handful of allocations, however many
values it has.
The parser accepts exactly what json::parse accepts, with every
combination of ignore_comments and ignore_trailing_commas, with and
without a trailing NUL, and under JSON_STRICT_NUL_HANDLING. It is
portable C++11 and does not depend on byte order.
basic_json_document adds parse, parse_copy, accept, read (reuses a
document's memory), root, is_discarded, source, owns_source,
node_count, memory_usage, and shrink_to_fit. basic_json_view adds
type, the is_* queries, operator bool, size, empty, materialize,
and source_offset. A parse error throws the same exception
basic_json::parse would throw for the same input, message and
position included.
detail::abi_config keeps JSON_STRICT_NUL_HANDLING and
JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON readable after json.hpp
undefines them, in the ABI namespace so they always match the
basic_json in use.
A NUL byte that ends a // comment is the end of the input, as in
parse() since #5696.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Fixes every CI failure develop has had since the API performance stack (#5831-#5842) was merged: compiler warnings and errors (Clang, GCC 4.9/11, MSVC 2015/2017/2022), clang-tidy, cpplint, Cppcheck, IWYU, the examples check, the no-exceptions, diagnostic-positions, implicit-conversions and strict-NUL test configurations, and the MinGW link (now with LLD).
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
These PRs were merged while their CI runs were cancelled, so the warnings
went unnoticed. All of them fail the -Werror builds:
- json_pointer.hpp (#5836): size_t += int under GCC -Wsign-conversion
- binary_reader.hpp (#5839): char_int_type passed as char under GCC
-Wconversion
- json.hpp (#5838): @throw without a description under clang
-Wdocumentation
- string_escape.hpp (#5836): escape_into() was inserted between escape()
and its documentation, so the @return was attached to a void function
- unit-diagnostics.cpp (#5832): clang -Wnrvo in a lambda without an
explicit return type
- unit-to_json-rvalue.cpp (#5837): GCC -Wnoexcept (C++20) and clang
-Wunneeded-member-function on the test type
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
MSVC 2015 warns C4554 on the shifts with casts in the fractional part
of zmij::to_shortest despite the parentheses; with /WX every target
fails to build. Name the shift amounts as constants instead.
The shortest-digits test built the candidates with one digit fewer via
a stream with std::setprecision(digits.size() - 2); the standard library
of MSVC 2017 ignores a precision of 0, so for two-digit values the
"shorter" candidates were the value itself. Truncate the digits
instead: the value lies within half a unit of the last digit, so the
truncated digits and their successor are the closest candidates.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Escape characters below 0x60 from a table straight into the write buffer
When the bulk scanner in dump_escaped_impl() stops at a control
character, a quote or a backslash, that byte used to go through the
UTF-8 decoder and a switch that wrote the escape byte by byte into
string_buffer, which was then copied into write_buffer at the next run.
A byte below 0x60 at a character boundary is a complete code point, so
it now skips the decoder: its escape sequence comes from a 96-entry
table of 8-byte rows (sequence padded with zeros, length in the last
byte), and all 8 bytes are copied straight into write_buffer with one
room check. The check sits inside the existing character-boundary
block, so it runs once per scanner stop and not once per byte; testing
every byte cost about 6% on twitter.json with ensure_ascii.
The switch in the decoder's accept branch lost its cases for
characters below 0x60, which can no longer reach it.
Output is byte-identical. The string-buffer boundary tests now use
\u-escaped non-ASCII characters, since simple escapes no longer reach
string_buffer, and a new test case checks every ASCII character at
every position against a reference escaper.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Avoid raw string literals with backslashes in CHECK macros
MSVC's traditional preprocessor mis-lexes raw string literals inside
macro arguments and reports illegal escape sequences (C2017, C3850).
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Sweep invalid UTF-8 at every position through dump()
Compare dump() of strings with ill-formed UTF-8 (stray continuation
bytes, invalid lead bytes, overlong encodings, surrogates, code points
above U+10FFFF, and sequences cut short by ASCII, a quote, a backslash,
a lead byte, or the end of the string) with an independent reference
serializer, for every error handler and with ensure_ascii on and off.
The pattern is put at every position of short strings to cross the
strides of the bulk scanners, and of longer ones to cross the end of the
512-byte string buffer.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
---------
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Fuzz the streaming input adapters against the contiguous one
The fast paths for contiguous input (bulk string scanning, multi-byte
loads, lazy position tracking) are taken only when parsing from a
pointer range; streams and non-contiguous iterators still read byte by
byte. The fuzzer drivers parsed only contiguous input, so a difference
between the two would ship unnoticed.
check_input_adapters() in fuzzer_common.hpp reads the input through a
std::istringstream, a std::list iterator range, and (on POSIX) a FILE*
from fmemopen, and asserts the same value, or the same exception id,
message and byte position, as for the pointer range. All seven drivers
call it; the JSON driver checks parse and accept, each with and without
ignore_comments. It replaces the stream check of the BON8 driver.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Rename the reader parameter to avoid a flawfinder false positive
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
---------
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Allow the same type for number_integer_t and number_unsigned_t
basic_json did not compile with NumberUnsignedType == NumberIntegerType
because overloads on both types collided. Such a basic_json now stores
every integer as number_integer: the colliding overloads take a
placeholder type, unsigned C++ integers are converted to number_integer_t,
and value_t::number_unsigned maps to number_integer.
Also fixes two problems this configuration would hit: the unsigned
integer parser accepted values that wrap around to negative numbers, and
dump() overflowed when computing the absolute value of the minimum.
A repeated NumberFloatType gets a static_assert.
Refs #2573
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Use json::with_integers_t in the test and the docs
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* ✅ skip patch test on GCC 5 to avoid internal compiler error
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Fix CI: name the unsigned overload type through template parameters
basic_json used number_unsigned_overload_t<basic_json> while the class was
still incomplete. MSVC rejected it as a recursive alias declaration and GCC
4.9, 5 and 7 crashed with an internal compiler error. basic_json now uses
number_unsigned_overload_type<NumberIntegerType, NumberUnsignedType>, which
also makes the GCC 5 workaround in the patch test unnecessary.
Infer reported a null dereference for the number_unsigned case in
from_json, which cannot happen for a basic_json whose number_unsigned_t is
number_integer_t; assert it.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Address review: conjunction/negation trait, astyle-friendly alias, docs table order
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
---------
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
patch_inplace() built its error messages inline. In particular get_value()
concatenated the context-sensitive "operation 'x'" string on every call, even
when nothing was wrong. The patch exceptions now come from NEVER_INLINE
helpers: the index and key errors reuse basic_json's existing
throw_array_index_out_of_range/throw_key_not_found, and the seven
patch-specific messages get one helper each in detail/exceptions.hpp. Ids
and messages are unchanged, and unit-throw_messages pins them (with and
without JSON_DIAGNOSTICS).
citm_catalog patch: -6.6 % (Apple clang), -9.5 % (GCC 16).
The other remaining throw sites (parser::syntax_error, binary_writer,
serializer, input_adapters, enum_key_duplicate, format_error) were prototyped
and measured: no gain beyond noise, so they stay as they are.
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Speed up container destruction, iterator access and diff()
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
* Destroy the first levels of a value by recursion
The pointer-reversing walk in destroy_container() avoids the call stack but
moves a value around for every container it visits. Recurse for the first
recursion_depth_limit() levels instead - no allocation, no thread-local, a
bounded frame - and hand anything deeper to the walk. Destroy twitter
0.49 -> 0.36 ms, citm 0.80 -> 0.74 ms (3.12.0: 0.47 / 1.00).
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
---------
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Niels Lohmann