wt/json
1
0 Fork 0
mirror of https://github.com/nlohmann/json.git synced 2026-11-09 16:21:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs: add the blog post "How nlohmann/json 3.13.0 got 3× faster" (#6022)

Browse Source 
* docs: add the blog post "How 3.13.0 got faster"

A post for the documentation (Home > Blog) on how 3.13.0 became faster:
eight themes with the pull requests behind them, the ideas taken from other
libraries, what did not work, how the fast paths are tested, and the
compile-time cost. Every number comes from a run on 2026-10-27 (3.12.0 against
develop plus the open non-draft PRs, Apple M1 Max) or from the linked PR.

- docs/mkdocs/docs/home/blog/how_3_13_0_got_faster.md: the post.
- docs/mkdocs/scripts/figures/perf_blog.py: draws its 12 figures (overview,
  eight timeline panels, the SWAR test, the cliffs, json_view throughput) as
  inline SVGs in the nj-charts style; data/perf_timeline.csv holds the
  timeline.
- tests/benchmarks/history: history.cpp and history.py time one fixed program
  against the header of every commit that changed it, interleaved, and write
  one CSV row per commit with its PR. The timeline of the post comes from it.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: compare json_document with Glaze, sonic, RapidJSON and Rust parsers

Extends the comparison in "How 3.13.0 got faster" from yyjson, simdjson and
Boost.JSON to eleven engines: sonic-cpp, RapidJSON and Glaze (generic DOM,
lazy_json, reading into structs) in C++, sonic-rs, simd-json and serde_json in
Rust. Measured on an Apple M1 Max (Apple clang 21, Rust 1.99) and an x86-64
Haswell server (GCC 13, Clang 18). Every engine must return the same values,
string bytes and bit-identical number sums before it is timed.

The post now answers whether json_document is the fastest parser: it is the
fastest document parser at parse+dump everywhere and at parsing on Apple
silicon, ties with simdjson's DOM on x86-64, and loses to simdjson On-Demand
and Glaze structs when only a few fields are read.

tests/benchmarks/json_view/others holds the programs (bench_all.cpp, the Rust
crate with its Cargo.lock), summarize.py and the pinned versions and checksums.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: captions, test-file table and figure fixes for the 3.13.0 post

- caption every table and figure
- describe what makes twitter.json, citm_catalog.json, canada.json and
  jeopardy.json hard, and what each timeline panel exercises
- split the json_view layout figure into the tree and the index, with
  the example text above; fix the arrows into the string boxes
- arrows and a "less is better" axis label in the cliffs figure
- name nlohmann/json in the comparison tables and chart
- licenses of the adapted code in the thanks section

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: byline and licenses for the 3.13.0 post

- author and date below the title, styled as .nj-byline; its link gets
  no external-link icon
- the license of every library in the thanks section, checked against
  each project's license files; credit sonic-rs for QUOTE_TAB, which
  orjson adapted from it

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: dump comparison and split acknowledgments in the 3.13.0 post

- a dump workload in bench_all and bench_rs: each engine writes a
  document parsed beforehand, and every output must read back with
  bit-identical numbers; an optional argument selects one workload
- the post's comparison gains a write table, a throughput chart, a dump
  row for all three configurations, and the x86-64 cases it loses
- the thanks section is split into adapted code and borrowed ideas

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: x86-64 dump numbers with #5908's branch-free token digits

The dump of canada.json with GCC on x86-64 lost to yyjson by 0.78 because
GCC compiled #5908's token-digit selections as branches; 9fad526888 fixed
that. The post's x86-64 dump and parse+dump rows of canada.json are now
measured with it (GCC 0.95, Clang 0.91 against yyjson), theme 5 tells the
story, and "How we measured" says which rows include the commit.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Compare JSON libraries by use case in the json_view harness

bench_all and bench_rs now time use cases (validate, sum, select, edit,
minify, write) instead of library functions. Every library may solve a
use case in every way it offers; each variant's result is checked, and
three malformed inputs give its validation level (full, no UTF-8,
partial). summarize.py reports each library's fastest fully validating
variant and a faster one at a lower level, as text and as JSON.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: restructure the 3.13.0 post around themes and use cases

- deeper technical detail and code in the eight themes; pull request
  numbers move to the appendix, timeline steps are labelled by theme
- comparison with other libraries by use case, as heatmaps per
  platform, with validation levels and the caveats next to the results
- appendix: test files, all 182 suite rows with a histogram, timeline
  steps, algorithmic fixes, per-file tables, behaviour changes, setup,
  pull requests by theme

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Measure each library's fastest modes in the json_view harness

Address the points of a review from the other libraries' side:

- simdjson 5.0.8 (On-Demand structs, key selectors) and Glaze 9.1.0
- number-skipping modes: yyjson and RapidJSON raw numbers, Boost.JSON
  number_precision::none
- reused parsers and allocators for yyjson, RapidJSON and Boost.JSON
- Boost.JSON parse_into, Glaze generic with double numbers, and the
  same struct types for nlohmann, Glaze, simdjson and serde
- minified copies of twitter.json and citm_catalog.json
- a fourth probe (trailing garbage) and Glaze "(strict)" variants
- a new random variant order in every round, as a variant's time could
  depend on the one before it by up to 15 % on the Mac
- --files for other test files and --conformance for JSONTestSuite,
  with conformance-rs.sh for the Rust program

summarize.py also reports each library's best reused variant.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: rerun the library comparison of the 3.13.0 post with fairer rules

The figures and Appendix E use the reruns with each library's fastest
modes. Skipping number conversion while reading is now allowed for
every library; copying number text is still not chosen when writing.

The text now says who set the rules, where libraries that check less
win select, how minified files change the picture, and that the
reused jeopardy.json claim holds with every library reused. A new
section covers 19 held-out files and JSONTestSuite, including where
json_document loses, and lists what was not measured. Every
comparison number was recomputed; Glaze is credited in the intro.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: measure the 3.13.0 post against develop with #5964 and #5908

The suite (Appendix B, the short-version table, the overview and distribution
figures) and the timeline endpoint now measure develop d20c4dcd89 with #5964
and #5908 merged. The geometric mean stays 2.9x (2.95); Dump/signed_ints and
Dump/unsigned_ints gain 1.5x and 2.1x from #5964; GetVectorDouble/floats drops
from 10.3x to 7.0x because the old candidate's suite binary happened to compile
that loop faster (a standalone build takes 0.97 ms with both headers).

The timeline endpoint is the 10-27 value scaled by the ratio of the new to the
old candidate, both measured on 10-30, so that it stays comparable with the
other points of the timeline.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: add three figures to the 3.13.0 post

- short strings: two overlapping 16-byte loads and stores
- eight digits at a time: the three multiplications, lane by lane
- validate on the 19 held-out files, per file, Mac and x86-64

The prose the figures replace is trimmed.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: measure the 3.13.0 post against develop with all open pull requests

The baseline is now develop 9c349b1d7c with every open non-draft pull
request merged, including #5968, #5970, #5971, #5975, #5976, #5978,
#5979, #5980, #5981 and #5982. The suite was rerun on 11-01 with every
step repeated when the load rose above 5 while it ran; the comparison
with other libraries and the held-out files were rerun on 10-31.

- suite: 3.0x geometric mean (median 2.9x); Appendix B, the overview
  and the distribution regenerated
- timeline endpoint, Appendix D and the cliffs figure updated
- the held-out figure compares json_document with the fastest other
  library
- new paragraphs for the pull requests listed above

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: measure the 3.13.0 post against develop with the open pull requests of 11-02

Suite, timeline endpoint, Appendix D and the scalar-run numbers now come from
develop b2674a4a2f with #5990, #5991, #5992, #5996, #5997, #5998, #5999 and
#6000 merged. The scalar runs are measured against a build without them on
develop's code, which no longer checks for a scalar before entering the loop.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: measure the 3.13.0 post against develop with all open pull requests of 11-02

Suite, timeline endpoint, Appendix D and the scalar-run numbers now come from
develop b2674a4a2f with #5990-#5992, #5996-#6006 and #6008 merged. With #6008,
arrays of one element skip the scalar run, so nested arrays no longer pay for
it. The suite ran three rounds per version; the round of 3.12.0 and of 3.13.0
that differed from the other two was dropped, as the setup table says.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: fix the facts the review of the 3.13.0 post found wrong or imprecise

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: re-measure the 3.13.0 post and apply the review

Measure everything again against develop with the open pull requests of
11-03 (suite, timeline, comparison on the Mac and on x86-64, compile
times, stream gap), add the suite on an Intel MacBook Pro as a check,
mark the ways that keep number text as "N" in the use-case charts, move
the loop-shape box to theme 5, and apply the style and wording fixes of
the review.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: make the comparison of the 3.13.0 post fairer

Three reviews of "How it compares" (a maintainer audit of the harness, a
methodology review, a reader's red team) led to a full rerun on the Mac and
x86-64 with simdjson 5.1.1 and new ways in the harness:

- in-place parses with the copy made before the timer ("(copy untimed)",
  never chosen as a library's best) and json_document::parse_copy();
- RapidJSON through StringStream, and with default precision in validate,
  which reads no number;
- BENCH_ONLY takes several prefixes.

The post now gives the copy's real cost (11-14 us for twitter.json on the
Mac, not 30), what yyjson gains without it, the run-to-run noise, the LTO
check, the single-file losses that were missing, and qualifies the headline
claim. Every number in the section and Appendix E is recomputed; the
NUL-byte remark moved out of the body.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Format bench_all.cpp with astyle

The amalgamation check runs astyle over tests/ and failed on bench_all.cpp. The variant tables keep one variant per
line between *INDENT-OFF* and *INDENT-ON*; astyle formats the rest, and the list of use cases becomes a named array so
the loop header stays on one line.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Print the history benchmark's results with a stream

Flawfinder flagged the printf in history.cpp (CWE-134), although its format string is a constant. The stream prints the
same text, so history.py reads it as before.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: address review comments on the performance blog post

- fix wording: "are new", "for each input source", "the strings", "the last zero to seven bytes", "by" Danila
  Kutenin, "the result of that query"
- introduce `json_document` where it first gets a detail, say that the `-mbranches` flag is not used, explain
  why the position and token bookkeeping overloads are empty, and say that a step's percentage is relative to
  the time before the step and that a median of 4 is the mean of the two middle values
- write percentages without a space before `%` in the post and in the price figure (regenerated)

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: wrap the prose of the performance blog post at 120 columns

Only line breaks change: the rendered page is identical after whitespace normalization.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: blog post: address the second round of review comments

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: blog post: note that Boost.JSON's tree lead comes from its arena

The comparison builds Boost.JSON's tree in a monotonic_resource. With its
default allocator and precise numbers, boost::json::parse() takes as long
as json::parse() on twitter.json, 5% longer on citm_catalog.json and 17%
less time on canada.json (M1, C++23 -O3, best of 3 x 30 runs; the arena
saves Boost.JSON 28-41%).

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: blog post: comparing already passes the depth; copying measured slower without the counter

operator== walks the standard containers itself and passes the depth as an
argument (#5840). A copy that does the same, inserting element by element,
measured 4-16% slower on an M1 Max with Apple clang, so copying keeps the
thread_local counter.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* docs: blog post: name every user of the depth counter and every input the document copies

The "Bounded recursion" paragraph said only copying and equality for custom
containers use the thread_local counter. Converting from another basic_json
specialization and the ordering comparisons use it too; equality for the
standard containers passes the depth instead.

The "Strings stay in the input" bullet named only parse_copy() and an rvalue
std::string as owning inputs. Streams, FILE* and other input that is not one
contiguous block of bytes are read into an owned buffer as well.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

---------

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
This commit is contained in:
Niels Lohmann authored and GitHub committed 2026-11-09 08:11:10 +01:00
1 parent 6343775000
commit 2c7474affc
51 files changed
+73792 -18
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

tests/benchmarks/json_view/others/summarize.py
Executable
+229
View File
@@ -0,0 +1,229 @@
#!/usr/bin/env python3
# __ _____ _____ _____
# __| | __| | | | JSON for Modern C++ (supporting code)
# | | |__ | | | | | | version 3.12.0
# |_____|_____|_____|_|___| https://github.com/nlohmann/json
#
# SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
# SPDX-License-Identifier: MIT
"""Summarize runs of bench_all and bench_rs by use case.
Usage: summarize.py LOG... [--rust LOG...] [--json OUT.json]
Every variant carries a validation level from the malformed-input probes: "full" (rejects truncated input, a stray
comma and invalid UTF-8), "no UTF-8" (accepts only the invalid UTF-8), "partial" (accepts or crashes on a structural
error, or crashes on any probe), or "n/a" (write: no input to validate; counts as full). A variant's level is the
worst it got for any file. Prints, for every use case and file,
(a) per library the best "full" variant and, if faster, the best variant at any level, with their ratios to the
best "full" nlohmann json_document variant (variants named "(reused)" or "(reused allocator)" keep a parser,
document or allocator across calls and are never chosen as a library's best; neither are "(copy untimed)"
variants, which copy their input before the timer starts), and in a "reused:" line the library's best reused
variant in the same way,
(b) all variants with their levels and times,
(c) the variants below "full" with what the probes found, and the variants excluded for wrong results.
--json writes {use case: {file: {library: {"best": {"variant", "us", "level"} or null, "best_any": {...},
"best_reused": {...} or null, "best_reused_any": {...} or null, "variants": {variant: {"us", "level"}}, "excluded":
{variant: reason}}}}} for the chart scripts ("best_reused" is the best full reused variant, "best_reused_any" the best
reused variant at any level); variant names there omit the "<library>: " prefix.
Several logs are combined by taking each cell's best time. A variant excluded for one file (or in one log) is excluded
for every file of that use case. The reference values that bench_all and bench_rs print for sum and select are
compared, so both programs are known to have checked against the same numbers.
"""
import json
import re
import sys
USE_CASES = ["validate", "sum", "select", "edit", "minify", "write"]
DOCS = ["twitter", "citm_catalog", "canada", "jeopardy", "status", "twitter_min", "citm_catalog_min"]
OURS = "nlohmann json_document"
RANK = {"full": 0, "n/a": 0, "no UTF-8": 1, "partial": 2}
CELL = re.compile(r"^(.+) \[(full|no UTF-8|partial|n/a)\] ([\d.]+)us(?: \(([\d.]+)x\))?$")
REF = re.compile(r"^check (\S+)\s+(\S+)\s+reference: (.*)$")
def library(name):
return name.split(": ", 1)[0]
def variant(name):
return name.split(": ", 1)[1] if ": " in name else name
def is_reused(name):
return re.search(r"\(reused[ )]", name) is not None
def is_untimed_copy(name):
return "(copy untimed)" in name
def worse(a, b):
return a if a is not None and (b is None or RANK[a] >= RANK[b]) else b
def parse(path, cells, levels, notes, excluded, refs):
for line in open(path, encoding="utf-8"):
line = line.rstrip("\n")
if line.startswith(("excluded: ", "level: ")):
kind, rest = line.split(": ", 1)
fields = rest.split(" | ")
doc, uc = fields[0].split()
name = fields[1]
target = excluded if kind == "excluded" else notes
for why in fields[-1].split("; "):
target.setdefault((uc, name), {}).setdefault(why, set()).add(doc)
if kind == "level":
levels[(uc, name)] = worse(levels.get((uc, name)), fields[2])
continue
m = REF.match(line)
if m:
refs.setdefault((m[1], m[2]), {})[path] = m[3]
continue
if line.startswith(("check", "done", "WARNING")) or " | " not in line:
continue
parts = line.split(" | ")
head = parts[0].split()
if len(head) != 2 or head[1] not in USE_CASES:
continue
for c in parts[1:]:
m = CELL.match(c.strip())
if m:
key = (head[0], head[1], m[1])
if head[0] not in DOCS: # bench_all --files: the documents are the files of a directory
DOCS.append(head[0])
cells[key] = min(cells.get(key, float("inf")), float(m[3]))
levels[(head[1], m[1])] = worse(levels.get((head[1], m[1])), m[2])
def check_refs(refs):
"""The sum and select references of all logs must agree (bench_rs prints Rust's float formatting)."""
def numbers(text):
return [float(x) for x in re.findall(r"[-+]?\d[\d.]*(?:e[-+]?\d+)?", text)]
ok = True
for (doc, uc), by_log in sorted(refs.items()):
values = {tuple(numbers(t)) for t in by_log.values()}
if len(values) > 1:
ok = False
print("WARNING: the %s %s references differ: %s" % (doc, uc, "; ".join("%s: %s" % kv for kv in by_log.items())))
return ok
def is_full(level):
return RANK[level] == 0
def main():
args = sys.argv[1:]
out_json = None
if "--json" in args:
i = args.index("--json")
out_json = args[i + 1]
del args[i:i + 2]
cpp = args[:args.index("--rust")] if "--rust" in args else args
rust = args[args.index("--rust") + 1:] if "--rust" in args else []
if not cpp and not rust:
sys.exit(__doc__)
cells, levels, notes, excluded, refs = {}, {}, {}, {}, {}
for p in cpp + rust:
parse(p, cells, levels, notes, excluded, refs)
check_refs(refs)
result = {}
def docs_of(uc):
return [d for d in DOCS if any(k[0] == d and k[1] == uc for k in cells)]
def cell_of(doc, uc):
return {n: t for (d, u, n), t in cells.items() if d == doc and u == uc and (uc, n) not in excluded}
def show(n, t, ours, uc):
ratio = "%5.2f" % (t / ours) if ours else " -"
return "%-40s %-10s %11.1f us %s" % (variant(n), "[%s]" % levels[(uc, n)], t, ratio)
print("# (a) Per library: the best full variant, and the best at any level if faster (ratio to the best full %s"
" variant; below 1: faster)" % OURS)
for uc in USE_CASES:
if not docs_of(uc):
continue
print("\n== " + uc)
for doc in docs_of(uc):
cell = cell_of(doc, uc)
libs = {}
for n, t in cell.items():
libs.setdefault(library(n), {})[n] = t
entry = result.setdefault(uc, {}).setdefault(doc, {})
rows = []
def pick(vs):
"""the best full variant and the best variant at any level"""
full = {n: t for n, t in vs.items() if is_full(levels[(uc, n)])}
return (min(full.items(), key=lambda x: x[1]) if full else None,
min(vs.items(), key=lambda x: x[1]) if vs else None)
def described(x):
return {"variant": variant(x[0]), "us": x[1], "level": levels[(uc, x[0])]} if x else None
for lib, vs in libs.items():
b, a = pick({n: t for n, t in vs.items() if not is_reused(n) and not is_untimed_copy(n)})
rb, ra = pick({n: t for n, t in vs.items() if is_reused(n)})
entry[lib] = {
"best": described(b),
"best_any": described(a),
"best_reused": described(rb),
"best_reused_any": described(ra),
"variants": {variant(n): {"us": t, "level": levels[(uc, n)]} for n, t in sorted(vs.items(), key=lambda x: x[1])},
}
rows.append((lib, b, a, rb, ra))
for (u, n), whys in excluded.items():
if u == uc:
lib = entry.setdefault(library(n), {"best": None, "best_any": None, "best_reused": None, "best_reused_any": None, "variants": {}})
lib.setdefault("excluded", {})[variant(n)] = "; ".join(sorted(whys))
ours = next((b[1] for lib, b, a, rb, ra in rows if lib == OURS and b), None)
print("%s%s" % (doc, "" if ours else " (no full %s result)" % OURS))
rows.sort(key=lambda r: (r[1] is None, r[1][1] if r[1] else r[2][1] if r[2] else 0))
for lib, b, a, rb, ra in rows:
line = " %-24s %s" % (lib, show(*b, ours, uc) if b else "%-40s %-10s %14s %5s" % ("(none)", "", "", ""))
if a and (not b or a[0] != b[0]):
line += " | faster: %s" % show(*a, ours, uc)
print(line)
if ra:
line = " %-22s %s" % ("reused:", show(*rb, ours, uc) if rb else "%-40s %-10s %14s %5s" % ("(none)", "", "", ""))
if not rb or ra[0] != rb[0]:
line += " | faster: %s" % show(*ra, ours, uc)
print(line)
print("\n# (b) All variants")
for uc in USE_CASES:
if not docs_of(uc):
continue
print("\n== " + uc)
for doc in docs_of(uc):
cell = cell_of(doc, uc)
ours = min((t for n, t in cell.items() if library(n) == OURS and not is_reused(n) and is_full(levels[(uc, n)])), default=None)
print(doc)
for n, t in sorted(cell.items(), key=lambda x: x[1]):
ratio = "%5.2f" % (t / ours) if ours else " -"
print(" %-66s %-10s %11.1f us %s" % (n, "[%s]" % levels[(uc, n)], t, ratio))
print("\n# (c) Variants below full, and excluded variants")
for uc in USE_CASES:
rows = sorted((n, whys) for (u, n), whys in notes.items() if u == uc)
ex = sorted((n, whys) for (u, n), whys in excluded.items() if u == uc)
if not rows and not ex:
continue
print("\n== " + uc)
for n, whys in rows:
for why, docs in sorted(whys.items()):
print(" %-58s %-10s %s (%s)" % (n, "[%s]" % levels.get((uc, n), "?"), why, ", ".join(d for d in DOCS if d in docs)))
for n, whys in ex:
for why, docs in sorted(whys.items()):
print(" %-58s %-10s %s (%s)" % (n, "excluded", why, ", ".join(d for d in DOCS if d in docs)))
if out_json:
with open(out_json, "w", encoding="utf-8") as f:
json.dump(result, f, indent=2)
f.write("\n")
if __name__ == "__main__":
main()