Profiling tl_2022_us_county.json (sample(1) on Apple Silicon) showed
~38% of parse time in allocator work and ~14% in std::string::push_back
during string-token construction. These changes target the low-hanging
fruit from that profile:
- Pre-reserve 2 slots in json_array and 4 slots in json_hash so
coordinate `[x, y]` pairs and typical GeoJSON property maps avoid
the 0 -> 1 -> 2 -> 4 vector-growth chain (and the shared_ptr copies
it incurs).
- Reuse a parser-wide std::string buffer for JSON_STRING tokens
instead of constructing a fresh local std::string per token. The
buffer is cleared (capacity preserved) at the start of each token
and copied into the final json_string, so once it has grown to the
longest string seen it stops reallocating entirely.
- std::move the freshly-created container shared_ptr into the parser
container stack in the `[` and `{` handlers, and move it out of the
frame on the matching `]` / `}`. Each move skips one atomic
inc/dec round-trip per container open and close.
On a tl_2022_us_county.json benchmark (4-iter user-time mean, Apple
Silicon, /usr/bin/time):
- main baseline: ~8.17s
- jsonpull-cpp before these changes: ~10.90s (+33%)
- jsonpull-cpp with these changes: ~9.33s (+14%)
So this commit recovers roughly half of the post-port regression.
The remaining gap is dominated by shared_ptr atomic refcount traffic
on the parse tree and per-node heap allocations, which would require
the larger unique_ptr/arena reworks to address.
Co-authored-by: Cursor <cursoragent@cursor.com>
Covers the `c` vs `ch` bug fixed in the previous commit: parsing
"\uD83D\uE000" (a valid high surrogate followed by a non-surrogate
BMP code point) used to mis-classify U+E000 as a low surrogate and
combine the two units into U+1F400 (F0 9F 90 80). The fixed code
flushes the stale high surrogate as standalone CESU-8 (ED A0 BD)
and then encodes U+E000 normally as EE 80 80. Verified the test
fails under the pre-fix logic.
Co-authored-by: Cursor <cursoragent@cursor.com>
- jsontool.cpp `out()`: route JSON_NUMBER (and anything else non-string)
through `json_stringify` instead of `o->string()`, which now asserts
on a non-string type and would crash `--extract` on numeric attributes.
- geojson.{hpp,cpp} `json_end_map`: take `json_pull_ptr` by reference so
the caller's shared_ptr is released, null-guard before touching
`jp->source`, and clear `jp->source` after delete to avoid a dangling
pointer.
- jsonpull/jsonpull.cpp: low-surrogate range check was comparing the
outer-loop byte `c` instead of the parsed code unit `ch`, breaking
surrogate-pair decoding for some \\uXXXX escapes. Pre-existing bug
preserved across the port.
- tile-join.cpp `handle_vector_layers`: require the field value to have
type JSON_STRING (and the key to be non-null) before calling
`string()`; the previous truthy `type` check would assert on a
non-string value.
Co-authored-by: Cursor <cursoragent@cursor.com>
json_number used to carry three parallel 8-byte fields (a double plus
both a 64-bit unsigned and a 64-bit signed slot for the large-integer
cases) even though at most one of the integer slots is ever the
canonical value for any given number. Collapse them into a
discriminated union:
enum repr_t { REPR_DOUBLE, REPR_LARGE_UNSIGNED, REPR_LARGE_SIGNED };
repr_t repr;
union { double d; unsigned long long u; long long s; } value;
Callers keep the same read API: number() returns the appropriate
double, large_unsigned() returns the ull (or 0 if not currently stored
that way), large_signed() likewise. Writes go through new set_number /
set_large_unsigned / set_large_signed methods that keep the
discriminator and the union value in sync.
This was prompted by an observation that moving json_type to the end
of the object should shrink things via tail-padding reuse. Empirically
the type-at-end rearrangement saves nothing on its own (every
subclass payload is 8-byte aligned so it can't slot into the 4-byte
tail), but the discriminated-number redesign hits the same idea from
a different direction: adding the 4-byte `repr` to json_number makes
the class non-standard-layout, which lets the Itanium ABI pack `repr`
into the base's 4-byte tail padding at offset 20. The union value
then starts at the natural offset 24, and json_number ends at offset
32 -- a 33% reduction.
Per-node sizes:
json_object (TRUE/FALSE/NULL) 24 bytes
json_number 32 bytes (was 48)
json_string 48 bytes
json_array 48 bytes
json_hash 48 bytes
Numbers dominate real GeoJSON (every coordinate is one), so the net
memory win on a typical parse is substantial.
Co-authored-by: Cursor <cursoragent@cursor.com>
`expect` was only meaningful while the parser was building a container,
and only ever read or written from jsonpull.cpp itself; once parsing
finished it was dead weight on every JSON_ARRAY and JSON_HASH (and
present-but-unused on every primitive too). Move it into the parser's
container stack, alongside the shared_ptr to the container it pertains
to:
struct json_pull::parse_frame {
json_object_ptr container;
json_type expect;
};
std::vector<parse_frame> container_stack;
The base class now only carries data-model state (parent, parser, type).
No external caller depended on `expect`, so no sweep was needed outside
jsonpull.cpp.
This change does not, in itself, shrink any json_object: the 4-byte
`expect` field used to live at offset 20 inside the base, where it was
already being eaten by alignment padding for the 8-byte-aligned first
member of every subclass (std::string, std::vector, double). The win is
in the data model, not the byte count -- the 4-byte hole is still
there, but it is now available for a future subclass whose first member
is small enough to slot into it.
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace the parallel std::vector<json_object_ptr> keys / values on
json_hash with a single std::vector<json_entry>, where json_entry is
a small {key, value} aggregate. This still preserves insertion order
(the property the parallel vectors were providing) but removes the
"keep two vectors in lockstep" pattern, and call sites can now use
range-for with structured bindings:
for (auto &[k, v] : o->entries()) { ... }
Side effects:
* sizeof(json_hash) drops from 72 to 48 bytes (one fewer vector
header), matching json_array.
* The keys() and values() accessors on json_object are replaced by a
single entries() accessor returning std::vector<json_entry>&.
* All call sites were swept from the old paired-index pattern
(`o->keys()[i]` / `o->values()[i]`) to entry-based access. Where the
original pattern relied on `nprop = 0` to short-circuit iteration on
a null or non-hash `properties`, the rewrite now guards the loop
explicitly with `if (o->type == JSON_HASH)` so that calling
entries() doesn't trip the asserting downcast.
Co-authored-by: Cursor <cursoragent@cursor.com>
The previous "every member in a struct" layout cost 168 bytes per
json_object, even for JSON_NULL / JSON_TRUE / JSON_FALSE nodes that
have no payload. Splitting json_object into a small base class plus
json_number / json_string / json_array / json_hash subclasses brings
each instance down to just the size of its actual contents:
json_object (base, TRUE / FALSE / NULL) 24 bytes
json_number 48 bytes
json_string 48 bytes
json_array (empty) 48 bytes
json_hash (empty) 72 bytes
Other size wins along the way:
* Drop enable_shared_from_this<json_object> (its embedded weak_ptr
was 16 bytes per node). json_pull now keeps an explicit
container_stack and the parser no longer needs to resurrect a
shared_ptr from a raw `parent` walk.
* Remove the unused `refcon` slot from the string variant.
* No virtual destructor: shared_ptr keeps the deleter from the
original std::make_shared<json_xxx> call, so destroying a
shared_ptr<json_object> still runs the right subclass dtor.
The base class exposes type-tagged accessors (o->string(),
o->number(), o->array(), o->keys(), o->values(), o->large_signed(),
o->large_unsigned()) that assert the type matches and downcast to
the appropriate subclass storage. All call sites were swept from
the old `o->value.X.Y` field paths to these accessors. A raw-pointer
overload of json_hash_get() replaces the few external uses of
shared_from_this() that survived in geojson-loop.cpp.
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace the manual malloc/realloc/free memory management in jsonpull
with std::shared_ptr ownership. Each json_object now owns its children
through std::vector<json_object_ptr>; raw back-pointers to parent and
parser remain valid by structural invariant and are cleared on
json_disconnect so detached subtrees can outlive their parser.
Strings become std::string, child arrays become std::vector, and the
old union becomes a struct so non-trivial members can coexist while
preserving the existing o->value.xxx access paths.
The old jsonpull.c is replaced by jsonpull.cpp, json_stringify now
returns std::string, and all callers across tippecanoe, tile-join,
tippecanoe-decode, tippecanoe-json-tool, tippecanoe-overzoom and the
unit tests are updated to use json_object_ptr / json_pull_ptr.
Co-authored-by: Cursor <cursoragent@cursor.com>
Enhance fqsort function to accept a temporary directory parameter for file handling. Update calls to fqsort in main.cpp, sort.cpp, sort.hpp, and unit.cpp to utilize the new parameter, ensuring temporary files are created in the specified directory.
* Divide-and-conquer polygon cleaning
* Catch the case where the gap can't be increased further
* Catch the case where we try to keep impossibly many features
* Make label points earlier in the tiling process
* Another case where it could try to drop even after already limiting.
* And do not coalesce on impossibly small geometries
* Add missing return
* Update version and changelog
* Improve the memory spike during tile construction
* Remove `need_tilestats`. Add a bunch of debug logging
* Remove debug logging
* Update version and changelog
* Add a flag to use an H3 index for the feature index
* Give mvt_value and serial_val a double-with-count concept
* Switch mean over to internal accumulation state
* Get rid of the attribute accumulation map
* Change vectors of features to vectors of pointers to features
* Fix --coalesce
* Revert "Add a flag to use an H3 index for the feature index"
This reverts commit b9b48f42c9.
* Update version and changelog
* Sketching out sqlite options for tile-join
* Enable sqlite3 serialized multithreading
* Fix some unnecessary round trips from std::string to char * and back
* Gathering join keys for sql query
* Make a query
* Actually open the gpkg. Fix the query quoting.
* Actually do the query and get results back
* Join the attributes onto the feature
* Set matched if the sql join matches
* Add a flag to get the feature ID from the query
* Observe attribute exclusion when joining from sql queries
* Add a flag to exclude all attributes from the tile side of the join
* Make tile-join bounding boxes reflect feature bounds, not tile bounds
* More tests
* An empty tileset has empty bounds at null island
* Gather the results from each thread *after* the thread finishes
* Missed a test
* Fix accidental inclusion of the top left of the tile in the bbox
* Case smashing and prefix trimming in the select
* Add test of sql join
* Make the join column option a join expression option
* Fix antimeridian adjustment. Z0 can't wait until the end of the tile
* Checkpoint on accepting multiple joined rows per tiled feature
* Adding the joined attribute should be per-feature, not per-attribute
* Forgot to update the test. Order of joined attributes has changed.
* Get the attributes back in the right order
* Add test of sql join with limit
* Allow multiple tile features to have the same join key
* Update tests for a country name with two distinct geometries
* Update version and changelog
* Forgot to mention the bounding box improvements
* Make tippecanoe-overzoom accept filters from a file
* Accept clip polygons from a file too
* Add test of clipping by polygon from file
* Add a test of reading an overzoom filter from a file
* Update version and changelog
* Plumb a clip bounding box around through overzoom
* Actually do some clipping
* Add a test
* Fix post-binning clipping
* Factoring out geometry parsing from feature parsing
* Accept a clip polygon argument to tippecanoe-overzoom
* Progress in the direction of polygon clipping
* Fix the wagyu flags. We need intersection, not union
* Remove debug spew
* Clip points to polygon bounds too
* Copy the geometric binning code to serve as intersection-finding code
* Add clipper2 for linestring clipping
* Compiles, but does not actually seem to clip. Hmm.
* Oh, it helps if I actually call the function
* Add clipping tests
* Add missing fixture, and don't crash if it is missing
* Remember to do polygon clipping after binning too
* Fix scaling before post-binning clipping. Add test.
* Remove unused parts of clipper
* Rename for consistency
* Revert accidentally added line
* Clip the clip regions to the tile bounds to reduce their complexity
* Add a test of clipping the clip region down to the tile boundary
* Update version and changelog
* Strip out unwanted attributes earlier in the process
* Skip aggregations whose attributes have been excluded
* Forgot one
* Add some more tests
* Update version and changelog
* Add a tippecanoe-decode option to restrict which attributes to decode
* Plumb buffer and feature limit around
* Check the feature limit
* Clarifying cases where output detail can be unspecified
* Clip bins to the tile buffer instead of just passing them through
* Add missing include
* Missed some tests
* Add --no-tile-compression option to tippecanoe-overzoom
* Update version and changelog
* Progress on plumbing a string pool for full_keys through
* More plumbing for key_pool
* Don't keep features with identical locations as multiplier features
* Revert "Don't keep features with identical locations as multiplier features"
This reverts commit 413f0c8024.
* Adjust calculated maxzoom to account for duplicate feature locations
* Update changelog and version
* Add a test affected by the maxzoom change with duplicate locations
* Round the drop rate a little for cross-platform test consistency
* Add an all-mvt_value attribute accumulation path
* Only bin by ID, not geometrically
* A little cleanup; changelog and version; test
* Remove accidental double-conversion
* Replace duplicated code with template
* Update changelog
* Choose the megatile features from those that will be in the next N zooms
* Take fractional zooms into account in multiplier feature choices
* Fix more tests
* Add a flag to retain multiplier features by minimum distance
* Limit feature expansion from multiplier density to 2x
* The multiplier cap was a bad idea
* Revert "The multiplier cap was a bad idea"
This reverts commit 6f8273a4c8.
* Revert "Limit feature expansion from multiplier density to 2x"
This reverts commit a26e41309d.
* Revert "Add a flag to retain multiplier features by minimum distance"
This reverts commit 01f14a4255.
* Remove the multiplier sequence, which should no longer matter
* Revert "Revert "Add a flag to retain multiplier features by minimum distance""
This reverts commit 776da4a1b8.
* Revert "Revert "Limit feature expansion from multiplier density to 2x""
This reverts commit 44a683d808.
* Revert "Revert "The multiplier cap was a bad idea""
This reverts commit 80f7cb1c0e.
* Track two kinds of previous index for next_feature
* Fix multiplier density threshold, I think
* Oh, I didn't git add the code changes
* Update version and changelog
* Try to install sqlite3 to fix the automated build
* Deleted too much
* Only let --preserve-point-density-threshold shift density around
* Remove the density debt concept, since it doesn't help
* Make the drop states a vector instead of an array
* Revert "Make the drop states a vector instead of an array"
This reverts commit 66c7abb6fa.
* Revert "Remove the density debt concept, since it doesn't help"
This reverts commit 707bb0c562.
* Revert "Only let --preserve-point-density-threshold shift density around"
This reverts commit ecf01f2231.
* Bin more aggressively if a point doesn't meet the pnpoly test
* Don't clip the points if we are binning
* Fix output of features added to the bin after its closure
* Update tests
* Update version and changelog
