Lazy feature collection#
The lazy vector type returned by the lazy vector reads.
| Module | pyramids_lazy._lazy_collection |
| Public as | pyramids.feature.LazyFeatureCollection |
| Extra required | parquet (pyarrow, dask-geopandas) |
What it is — and is not#
A LazyFeatureCollection is a dask_geopandas.GeoDataFrame. It is not a
FeatureCollection and not a geopandas.GeoDataFrame. Both it and the eager
FeatureCollection satisfy core's SpatialObject protocol, so a consumer accepts either through
isinstance(x, SpatialObject) rather than by checking for a concrete class.
That subclass relationship is why the whole dask-geopandas API is available directly on the object
— npartitions, spatial_shuffle(), compute(), the geometry accessors — rather than being
re-exposed method by method.
Why the import is deferred#
This module hard-imports dask_geopandas, which only the parquet extra installs. The class is
therefore resolved through a module-level __getattr__ on the package root, so importing
pyramids_lazy on a minimal install never evaluates this file. LazyNetCDF, which needs only
dask, is imported eagerly instead.
Pickling is handled explicitly: a reconstruction hook rebuilds the subclass rather than letting
dask's default reduce return a plain dask_geopandas.GeoDataFrame, so a round trip through a
distributed worker preserves the type.
API#
pyramids_lazy._lazy_collection.LazyFeatureCollection
#
Bases: GeoDataFrame
Lazy vector collection backed by a dask_geopandas graph.
Subclasses :class:dask_geopandas.GeoDataFrame — every method of that
parent is available unchanged (to_crs, clip, sjoin,
.npartitions, .spatial_partitions). Overrides: compute /
persist return pyramids-native wrappers, spatial_shuffle
re-wraps the shuffled result, to_file / plot raise with
actionable messages, and read_file is a classmethod delegating
back to the eager reader so the
:class:~pyramids.base.protocols.SpatialObject protocol is satisfied.
Construct via FeatureCollection.read_file(..., backend='dask') or
FeatureCollection.read_parquet(..., backend='dask'). Direct
construction is discouraged; use
:meth:LazyFeatureCollection.from_dask_gdf if you must wrap a
pre-built dask-geopandas frame.
Note
On installs without the [parquet] extra, importing this
class from :mod:pyramids.feature returns None (sentinel).
Consumer code using isinstance must guard explicitly:
if LazyFeatureCollection is not None and isinstance(x, LazyFeatureCollection):....
Source code in src/pyramids_lazy/_lazy_collection.py
60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 | |
epsg
property
#
EPSG code of the CRS, or None when the CRS is unset.
Cheap — reads the cached pyproj CRS attached to the
dask-geopandas frame and calls .to_epsg(). No graph
materialisation.
Examples:
- Inspect the EPSG of a lazy FC built from a WGS84 GDF:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1]}, ... geometry=[Point(0, 0)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=1) ... ) >>> lfc.epsg 4326
__getattribute__(name)
#
Return a class-swapped view of every inherited method's result.
dask-geopandas inherits from dask-expr and does not honour pandas'
classic _constructor hook. Methods like to_crs, clip,
sjoin, copy, drop_duplicates go through
map_partitions with meta inference and produce a plain
:class:dask_geopandas.GeoDataFrame, silently dropping the
:class:LazyFeatureCollection subclass. This hook wraps every
public callable so that a :class:dask_geopandas.GeoDataFrame
result is rebranded back to the subclass before it reaches the
caller. Private names (_-prefixed) are passed through
unchanged to keep dask-expr internals — meta inference, name
caches, partition probing — on the hot path they expect.
Source code in src/pyramids_lazy/_lazy_collection.py
__reduce__()
#
Preserve the LazyFeatureCollection class across pickle round-trips.
from_dask_gdf uses result.__class__ = cls to rebind the
concrete class. The parent's default __reduce__ reconstructs
with the base :class:dask_geopandas.GeoDataFrame, silently
dropping the subclass on unpickle. Wrapping super's reduce tuple
in a module-level trampoline re-applies the class swap.
Source code in src/pyramids_lazy/_lazy_collection.py
__repr__()
#
Pyramids-branded repr; avoids the generic Dask DataFrame one.
compute(**kwargs)
#
Materialize the graph and return an eager :class:FeatureCollection.
The from pyramids.feature import FeatureCollection
import below is inline by necessity — collection.py imports
this module from inside its dask branch, so a top-of-file import
here would form a cycle. Per CLAUDE.md, circular-import
breaks are the second accepted exception to the no-inline-imports
rule.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
**kwargs
|
Any
|
Forwarded to
:meth: |
{}
|
Returns:
| Name | Type | Description |
|---|---|---|
FeatureCollection |
FeatureCollection
|
an eager wrapper around the materialized |
FeatureCollection
|
class: |
Examples:
- Compute a two-partition lazy frame back to eager form:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import ( ... FeatureCollection, LazyFeatureCollection, ... ) >>> gdf = gpd.GeoDataFrame( ... {"v": [1, 2, 3, 4]}, ... geometry=[Point(i, i) for i in range(4)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=2) ... ) >>> eager = lfc.compute() >>> len(eager) 4 >>> eager.epsg 4326
Source code in src/pyramids_lazy/_lazy_collection.py
compute_total_bounds()
#
Materialise :attr:total_bounds via an explicit .compute() call.
:attr:total_bounds on a
:class:dask_geopandas.GeoDataFrame is a lazy dask Scalar.
Most consumers who want concrete numbers today wrote
lfc.total_bounds.compute() — this helper gives them a
single-line, self-documenting equivalent that also makes the
compute-cost explicit in the method name.
Returns:
| Type | Description |
|---|---|
Any
|
numpy.ndarray: |
Any
|
4-element 1-D array — the exact shape |
Any
|
class: |
Examples:
- Force the reduction and destructure into named values:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1, 2]}, ... geometry=[Point(0, 0), Point(10, 20)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=1) ... ) >>> xmin, ymin, xmax, ymax = lfc.compute_total_bounds() >>> float(xmin), float(ymax) (0.0, 20.0)
Source code in src/pyramids_lazy/_lazy_collection.py
from_dask_gdf(dask_gdf)
classmethod
#
Wrap an existing :class:dask_geopandas.GeoDataFrame as a LazyFC.
This is the canonical construction path for code that already holds
a pre-built :class:dask_geopandas.GeoDataFrame (for example,
from :func:dask_geopandas.from_geopandas or a third-party reader)
and wants to use it through the pyramids type system.
Implementation uses class-swap (copy() + __class__
assignment) rather than a four-arg reconstruction, because
(dask, _name, _meta, divisions) is private dask.dataframe API
that shifted between dask-expr and legacy. Class-swap relies only
on Python's object model and survives upstream churn. The
invariant "LazyFeatureCollection adds no extra instance state
beyond :class:dask_geopandas.GeoDataFrame" is pinned by
:func:tests.feature.test_lazy_feature_collection.test_from_dask_gdf_preserves_state.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
dask_gdf
|
Any
|
Pre-built |
required |
Returns:
| Name | Type | Description |
|---|---|---|
LazyFeatureCollection |
LazyFeatureCollection
|
A view over |
LazyFeatureCollection
|
|
|
LazyFeatureCollection
|
The task graph, partition names, metadata, and divisions |
|
LazyFeatureCollection
|
are unchanged — only the concrete class is swapped. |
Examples:
- Wrap a fresh dask_geopandas frame:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1, 2]}, ... geometry=[Point(0, 0), Point(1, 1)], ... crs="EPSG:4326", ... ) >>> ddf = dg.from_geopandas(gdf, npartitions=1) >>> lfc = LazyFeatureCollection.from_dask_gdf(ddf) >>> lfc.npartitions 1 >>> lfc.epsg 4326
Source code in src/pyramids_lazy/_lazy_collection.py
persist(**kwargs)
#
Persist partitions in worker memory; keep the FC lazy.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
**kwargs
|
Any
|
Forwarded to
:meth: |
{}
|
Returns:
| Name | Type | Description |
|---|---|---|
LazyFeatureCollection |
LazyFeatureCollection
|
A new lazy FC whose graph has been |
LazyFeatureCollection
|
materialized into worker memory. |
Examples:
- Persist and confirm the result is still lazy:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1, 2]}, ... geometry=[Point(0, 0), Point(1, 1)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=2) ... ) >>> persisted = lfc.persist() >>> isinstance(persisted, LazyFeatureCollection) True >>> persisted.npartitions 2
Source code in src/pyramids_lazy/_lazy_collection.py
plot(*args, **kwargs)
#
No lazy plot path; materialize first.
Raises:
| Type | Description |
|---|---|
NotImplementedError
|
Always. Call |
Examples:
- Calling
.plotdirectly raises with an actionable hint:>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1]}, ... geometry=[Point(0, 0)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=1) ... ) >>> try: ... lfc.plot() ... except NotImplementedError as exc: ... "compute().plot" in str(exc) True
Source code in src/pyramids_lazy/_lazy_collection.py
read_file(path, *, npartitions=None, chunksize=None, **kwargs)
classmethod
#
Construct a LazyFC from a vector file.
Satisfies the :class:~pyramids.base.protocols.SpatialObject
protocol's read_file classmethod requirement. Delegates to
:meth:FeatureCollection.read_file with backend='dask'.
The import of :class:FeatureCollection is inline by
necessity — :mod:pyramids.feature.collection imports this module
from inside its dask branch, so a top-of-file import here would
form a cycle. Per CLAUDE.md, circular-import breaks are an
accepted exception to the no-inline-imports rule.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path
|
str | Path
|
Path to a vector file GDAL/OGR can open (GeoJSON, Shapefile, GeoPackage, FlatGeobuf, …). |
required |
npartitions
|
int | None
|
Target number of dask partitions. Mutually
exclusive with |
None
|
chunksize
|
int | None
|
Target rows per partition. Mutually exclusive
with |
None
|
**kwargs
|
Any
|
Forwarded to :meth: |
{}
|
Returns:
| Name | Type | Description |
|---|---|---|
LazyFeatureCollection |
LazyFeatureCollection
|
A new lazy FC. |
Examples:
- Read a small GeoJSON as a lazy FC via the classmethod:
>>> import tempfile >>> from pathlib import Path >>> import geopandas as gpd >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1, 2, 3]}, ... geometry=[Point(i, i) for i in range(3)], ... crs="EPSG:4326", ... ) >>> d = Path(tempfile.mkdtemp()) >>> p = d / "pts.geojson" >>> gdf.to_file(p, driver="GeoJSON") >>> lfc = LazyFeatureCollection.read_file(str(p), npartitions=1) >>> lfc.npartitions 1 >>> len(lfc.compute()) 3
Source code in src/pyramids_lazy/_lazy_collection.py
spatial_shuffle(by='hilbert', level=16, *, npartitions=None, divisions=None)
#
Spatially sort rows via a space-filling curve.
Thin override of :meth:dask_geopandas.GeoDataFrame.spatial_shuffle
that rewraps the result as a :class:LazyFeatureCollection.
Populates :attr:spatial_partitions, enabling partition-pruned
:meth:sjoin on the result.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
by
|
str
|
One of |
'hilbert'
|
level
|
int
|
Curve resolution; default 16. |
16
|
npartitions
|
int | None
|
Optional output partition count. |
None
|
divisions
|
Any
|
Optional explicit division boundaries. |
None
|
Returns:
| Name | Type | Description |
|---|---|---|
LazyFeatureCollection |
LazyFeatureCollection
|
A new lazy FC with |
LazyFeatureCollection
|
populated. |
Examples:
- Shuffle a two-partition frame and confirm bboxes are set:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": list(range(8))}, ... geometry=[Point(i, i) for i in range(8)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=2) ... ) >>> shuffled = lfc.spatial_shuffle() >>> shuffled.spatial_partitions is not None True
Source code in src/pyramids_lazy/_lazy_collection.py
to_file(path, *args, **kwargs)
#
No lazy OGR write path exists; raise with an actionable message.
Raises:
| Type | Description |
|---|---|
NotImplementedError
|
Always. dask-geopandas has no lazy OGR
write path for GeoJSON / Shapefile / GeoPackage. Call
|
Examples:
- The hint at the raise site points to the supported paths:
>>> import geopandas as gpd >>> import dask_geopandas as dg >>> from shapely.geometry import Point >>> from pyramids.feature import LazyFeatureCollection >>> gdf = gpd.GeoDataFrame( ... {"v": [1]}, ... geometry=[Point(0, 0)], ... crs="EPSG:4326", ... ) >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=1) ... ) >>> try: ... lfc.to_file("out.geojson") ... except NotImplementedError as exc: ... "compute().to_file" in str(exc) True
Source code in src/pyramids_lazy/_lazy_collection.py
to_parquet(path, *args, **kwargs)
#
Write the lazy frame to GeoParquet (the one genuinely-lazy writer).
explicit wrapper around
:meth:dask_geopandas.GeoDataFrame.to_parquet. The parent
returns a :class:dask.delayed.Delayed (or None when the
call is implicitly computed), which leaks the dask detail into
user code and breaks the pyramids convention that to_*
writers return None. This override normalises the contract:
- The call blocks until every partition has been written.
- Return is always
None— matches :meth:FeatureCollection.to_parquetand :meth:LazyFeatureCollection.to_file(the raising stub). - All other kwargs (
compression=,write_index=,storage_options=) forward to the parent.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
path
|
Any
|
Destination path or directory. Cloud URLs ( |
required |
*args
|
Any
|
Forwarded to
:meth: |
()
|
**kwargs
|
Any
|
Forwarded to
:meth: |
{}
|
Raises:
| Type | Description |
|---|---|
ValueError
|
If |
Examples:
- Round-trip a lazy FC through GeoParquet:
>>> import tempfile # doctest: +SKIP >>> from pathlib import Path # doctest: +SKIP >>> import geopandas as gpd # doctest: +SKIP >>> import dask_geopandas as dg # doctest: +SKIP >>> from shapely.geometry import Point # doctest: +SKIP >>> from pyramids.feature import LazyFeatureCollection # doctest: +SKIP >>> d = Path(tempfile.mkdtemp()) # doctest: +SKIP >>> gdf = gpd.GeoDataFrame( ... {"id": [1, 2]}, ... geometry=[Point(0, 0), Point(1, 1)], ... crs="EPSG:4326", ... ) # doctest: +SKIP >>> lfc = LazyFeatureCollection.from_dask_gdf( ... dg.from_geopandas(gdf, npartitions=1) ... ) # doctest: +SKIP >>> lfc.to_parquet(d / "out.parquet") # doctest: +SKIP