pysits 2.0.0.dev1__tar.gz → 2.0.0.dev3__tar.gz

pysits-2.0.0.dev3/.claude/settings.local.json

@@ -0,0 +1,16 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(R --version)",
+      "Bash(pyenv versions *)",
+      "Bash(pyenv install *)",
+      "Bash(grep -E \"^  3\\\\.\\(10|11|12|13|14\\)\\\\.[0-9]+$\")",
+      "Bash(chmod +x run_matrix.sh)",
+      "Bash(./run_matrix.sh /Users/felipecarlos/.pyenv/versions/3.12.13/bin/python3 3.12)",
+      "Bash(Rscript *)"
+    ]
+  },
+  "disabledMcpjsonServers": [
+    "geo-infrastructure"
+  ]
+}

pysits-2.0.0.dev3/.github/workflows/tests.yaml

@@ -0,0 +1,96 @@
+name: Tests
+
+on:
+  workflow_dispatch:
+    inputs:
+      reason:
+        required: false
+        description: 'Reason'
+        default: 'Manual trigger'
+
+jobs:
+  test:
+    name: Python ${{ matrix.python-version }}
+    runs-on: ubuntu-latest
+
+    strategy:
+      # Report every version, so a single break does not hide the others.
+      fail-fast: false
+      matrix:
+        python-version: ['3.10', '3.11', '3.12', '3.13', '3.14']
+
+    env:
+      # Fail the build on R warnings that would otherwise pass unnoticed
+      R_KEEP_PKG_SOURCE: yes
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Set up R
+        uses: r-lib/actions/setup-r@v2
+        with:
+          use-public-rspm: true
+
+      - name: Install R dependencies
+        uses: r-lib/actions/setup-r-dependencies@v2
+        with:
+          # `arrow` and `kohonen` are required by pysits but are not part of a
+          # default `install.packages("sits")`, so they are listed explicitly.
+          packages: |
+            any::sits
+            any::arrow
+            any::kohonen
+            any::jsonlite
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Create a clean virtual environment
+        # Mirrors how users install pysits, so a broken dependency resolution
+        # is caught here rather than by them.
+        run: python -m venv .venv
+
+      - name: Install pysits
+        # `--only-binary` makes pip fail instead of falling back to a source
+        # build when no wheel matches this interpreter. Those builds have
+        # shipped binaries that segfault. Only pysits itself is built here.
+        run: |
+          .venv/bin/python -m pip install --upgrade pip
+          .venv/bin/python -m pip install \
+            --only-binary=:all: --no-binary pysits \
+            ".[dev]"
+
+      - name: Report resolved dependency versions
+        run: .venv/bin/python -m pip list
+
+      - name: Verify Arrow interoperability between R and Python
+        # pyarrow and R arrow package each bundle libarrow. If they stop
+        # sharing an allocator, transfers are silently corrupted rather than
+        # failing, so this asserts a known value survives a round-trip.
+        run: |
+          .venv/bin/python - <<'EOF'
+          import pyarrow as pa
+
+          import pysits
+          from pysits.conversions.tibble import (
+              pandas_sits_to_tibble,
+              tibble_sits_to_pandas,
+          )
+
+          probe = [0.5, 1.5, 2.5]
+          assert pa.array(probe).to_pylist() == probe, "pyarrow returns corrupted data"
+
+          samples = pysits.samples_modis_ndvi
+          expected = samples["time_series"][0]["NDVI"].tolist()
+          actual = tibble_sits_to_pandas(pandas_sits_to_tibble(samples.head(1)))
+          actual = actual["time_series"][0]["NDVI"].tolist()
+
+          assert actual == expected, f"round-trip corrupted data: {actual} != {expected}"
+          print("Arrow round-trip is lossless.")
+          EOF
+
+      - name: Run tests
+        run: .venv/bin/python -m pytest

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pysits
-Version: 2.0.0.dev1
+Version: 2.0.0.dev3
 Summary: Python wrapper for the sits R package
 Project-URL: Homepage, https://github.com/e-sensing/pysits
 Project-URL: Issue Tracker, https://github.com/e-sensing/pysits/issues
@@ -20,17 +20,26 @@ Requires-Dist: geopandas>=1.1.0
 Requires-Dist: jinja2>=3.1.6
 Requires-Dist: matplotlib>=3.10.1
 Requires-Dist: pandas>=2.2.3
+Requires-Dist: pandas>=2.3.3; python_version >= '3.14'
 Requires-Dist: pillow>=11.1.0
 Requires-Dist: pyarrow>=20.0.0
-Requires-Dist: rpy2<4.0.0,>=3.5.17
+Requires-Dist: rpy2<4.0.0,>=3.6.0
 Requires-Dist: shapely>=2.1.1
 Provides-Extra: dev
+Requires-Dist: affine>=2.4.0; extra == 'dev'
 Requires-Dist: cloudpickle>=3.1.1; extra == 'dev'
+Requires-Dist: dask>=2025.3.0; extra == 'dev'
+Requires-Dist: odc-stac>=0.5.3; extra == 'dev'
+Requires-Dist: planetary-computer>=1.0.0; extra == 'dev'
 Requires-Dist: pre-commit>=3.6.0; extra == 'dev'
+Requires-Dist: pystac>=1.15.0; extra == 'dev'
 Requires-Dist: pytest-cov>=6.1.1; extra == 'dev'
 Requires-Dist: pytest-randomly>=3.16.0; extra == 'dev'
 Requires-Dist: pytest>=7.4.4; extra == 'dev'
+Requires-Dist: rasterio>=1.3.0; extra == 'dev'
+Requires-Dist: rioxarray>=0.18.2; extra == 'dev'
 Requires-Dist: ruff>=0.11.12; extra == 'dev'
+Requires-Dist: xarray>=2025.3.0; extra == 'dev'
 Provides-Extra: xarray
 Requires-Dist: affine>=2.4.0; extra == 'xarray'
 Requires-Dist: dask>=2025.3.0; extra == 'xarray'

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/pyproject.toml

@@ -17,7 +17,7 @@
 
 [project]
 name = "pysits"
-version = "2.0.0.dev1"
+version = "2.0.0.dev3"
 description = "Python wrapper for the sits R package"
 readme = "README.md"
 requires-python = ">=3.10,<4"
@@ -58,8 +58,13 @@ classifiers = [
 ]
 
 dependencies = [
-    "rpy2>=3.5.17,<4.0.0",
+    # 3.6.0 is the first release distributing wheels. Earlier ones are
+    # source-only, so pip would compile rpy2 against the local R installation
+    "rpy2>=3.6.0,<4.0.0",
     "pandas>=2.2.3",
+    # pandas ships no wheels for Python 3.14 before 2.3.3. Without this floor
+    # pip builds pandas from source there, and the result segfaults.
+    "pandas>=2.3.3; python_version>='3.14'",
     "pillow>=11.1.0",
     "matplotlib>=3.10.1",
     "geopandas>=1.1.0",
@@ -90,6 +95,9 @@ dev = [
     "pytest-cov>=6.1.1",
     "pytest-randomly>=3.16.0",
     "cloudpickle>=3.1.1",
+    # The test suite covers the xarray exporters and reads rasters directly
+    "pysits[xarray]",
+    "rasterio>=1.3.0",
 ]
 
 [tool.ruff]

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/pysits/__init__.py

@@ -17,10 +17,13 @@
 
 """pysits module."""
 
+# Importing `backend.arrow` first selects the libarrow allocator, which must
+# happen before anything imports `pyarrow` or loads the R `arrow` package.
+# Keep this import above the others. See `pysits.backend.arrow`.
+from .backend import arrow as _arrow  # noqa: F401
 from .conversions.dsl.mask import MaskValue
 from .conversions.dsl.tuning import hparam
 from .settings import __version__
-from .sits.classification import sits_classify, sits_label_classification, sits_smooth
 from .sits.colors import (
     sits_colors,
     sits_colors_reset,
@@ -62,6 +65,9 @@ from .sits.data import (
     sits_apply,
     sits_bands,
     sits_bbox,
+    sits_classify,
+    sits_encode,
+    sits_label_classification,
     sits_labels,
     sits_labels_summary,
     sits_list_collections,
@@ -69,10 +75,17 @@ from .sits.data import (
     sits_mixture_model,
     sits_reduce,
     sits_select,
+    sits_smooth,
     sits_timeline,
 )
 from .sits.data import sits_summary as summary
-from .sits.exporters import sits_as_geopandas, sits_as_xarray, sits_to_csv, sits_to_xlsx
+from .sits.exporters import (
+    sits_as_geopandas,
+    sits_as_xarray,
+    sits_timeseries_to_csv,
+    sits_to_csv,
+    sits_to_xlsx,
+)
 from .sits.impute import (
     impute_linear,
     impute_mean,
@@ -87,6 +100,7 @@ from .sits.ml import (
     sits_kfold_validate,
     sits_lightgbm,
     sits_lighttae,
+    sits_lstm_fcn,
     sits_mlp,
     sits_model_export,
     sits_pre_train,
@@ -107,7 +121,6 @@ from .sits.ts import (
     sits_cluster_clean,
     sits_cluster_dendro,
     sits_cluster_frequency,
-    sits_encode,
     sits_geo_dist,
     sits_get_class,
     sits_get_data,
@@ -118,6 +131,7 @@ from .sits.ts import (
     sits_pred_references,
     sits_pred_sample,
     sits_predictors,
+    sits_random_sampling,
     sits_reduce_imbalance,
     sits_sample,
     sits_sampling_design,
@@ -125,6 +139,7 @@ from .sits.ts import (
     sits_som_clean_samples,
     sits_som_evaluate_cluster,
     sits_som_map,
+    sits_som_remove_samples,
     sits_stats,
     sits_stratified_sampling,
     sits_validate,
@@ -137,13 +152,21 @@ from .sits.utils import (
     read_rds,
 )
 from .sits.visualization import sits_plot as plot
-from .sits.visualization import sits_view
+from .sits.visualization import sits_sankey, sits_view
 
 __all__ = (
     # Classification
     "sits_classify",
     "sits_smooth",
     "sits_label_classification",
+    # Embeddings
+    "sits_pre_train",
+    "sits_encode",
+    "sits_ssl_mae",
+    "sits_ssl_lejepa",
+    "sits_ssl_vicreg",
+    "sits_barlow_twins",
+    "sits_contrastive_learning",
     # Cube
     "sits_cube",
     "sits_clean",
@@ -170,7 +193,7 @@ __all__ = (
     "sits_config_user_file",
     "sits_config_value",
     "sits_parallel",
-    # Data management
+    # Data
     "sits_bands",
     "sits_timeline",
     "sits_labels",
@@ -194,6 +217,7 @@ __all__ = (
     "sits_resnet",
     "sits_tempcnn",
     "sits_lighttae",
+    "sits_lstm_fcn",
     "sits_svm",
     "sits_xgboost",
     "sits_lightgbm",
@@ -202,12 +226,6 @@ __all__ = (
     "sits_model_export",
     "sits_formula_linear",
     "sits_formula_logref",
-    "sits_ssl_mae",
-    "sits_ssl_lejepa",
-    "sits_ssl_vicreg",
-    "sits_barlow_twins",
-    "sits_contrastive_learning",
-    "sits_pre_train",
     # Impute
     "impute_linear",
     "impute_mean",
@@ -228,13 +246,14 @@ __all__ = (
     "sits_som_map",
     "sits_som_evaluate_cluster",
     "sits_som_clean_samples",
+    "sits_som_remove_samples",
     "sits_geo_dist",
     "sits_patterns",
     "sits_sample",
     "sits_reduce_imbalance",
     "sits_sampling_design",
     "sits_stratified_sampling",
-    "sits_encode",
+    "sits_random_sampling",
     # Tiles
     "sits_tiles_to_roi",
     "sits_roi_to_tiles",
@@ -249,12 +268,14 @@ __all__ = (
     "sits_as_xarray",
     "sits_as_geopandas",
     "sits_to_xlsx",
+    "sits_timeseries_to_csv",
     # Tuning
     "sits_tuning_hparams",
     "sits_tuning",
     # Visualization
     "plot",
     "sits_view",
+    "sits_sankey",
     # DSL Variables
     "MaskValue",
     "hparam",

pysits-2.0.0.dev3/pysits/backend/arrow.py

@@ -0,0 +1,106 @@
+#
+# Copyright (C) 2025 sits developers.
+#
+# This program is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, see <https://www.gnu.org/licenses/>.
+#
+
+"""Arrow interoperability between R and Python.
+
+``pysits`` moves data between R and Python as Arrow IPC streams. That puts two
+independent builds of the libarrow C++ library in a single process: the one
+bundled in the R ``arrow`` package and the one bundled in ``pyarrow``. Their
+symbols collapse into a single namespace, so a call made through ``pyarrow``
+can be served by R's libarrow.
+
+Each build keeps a private ``mimalloc`` heap. When the two are mixed, a buffer
+allocated by one library is read back as zeroed memory by the other, and every
+transfer is silently corrupted instead of failing. Both builds also support the
+system allocator, which is shared, so selecting it keeps them interoperable.
+
+The allocator is chosen by libarrow the first time it initializes, so importing
+this module sets ``ARROW_DEFAULT_MEMORY_POOL``. It is imported before anything
+that pulls in ``pyarrow`` or the R ``arrow`` package.
+"""
+
+import os
+
+#
+# Allocator selected to keep R libarrow and pyarrow interoperable
+#
+ARROW_MEMORY_POOL_ENVVAR = "ARROW_DEFAULT_MEMORY_POOL"
+ARROW_MEMORY_POOL = "system"
+
+# Set on import: libarrow reads this once, when it initializes. A value already
+# present in the environment is left alone so users can override it.
+os.environ.setdefault(ARROW_MEMORY_POOL_ENVVAR, ARROW_MEMORY_POOL)
+
+#
+# Values transferred to R and back to verify the round-trip is lossless. A
+# mismatched pair of libarrow builds reads them back as zeros.
+#
+PROBE_COLUMN = "pysits_arrow_probe"
+PROBE_VALUES = [0.5, 1.5, 2.5]
+
+
+def check_arrow_memory_pool() -> None:
+    """Refuse to load R ``arrow`` package under an unshared allocator.
+
+    Called before the R ``arrow`` package is loaded. Corruption appears only
+    once both libarrow builds are in use, and is silent when it does, so the
+    configuration is rejected up front rather than probed for afterwards.
+
+    Raises:
+        RuntimeError: If the selected allocator is not the shared one.
+    """
+    selected = os.environ.get(ARROW_MEMORY_POOL_ENVVAR)
+
+    if selected == ARROW_MEMORY_POOL:
+        return
+
+    raise RuntimeError(
+        f"{ARROW_MEMORY_POOL_ENVVAR} is set to '{selected}', but pysits "
+        f"requires '{ARROW_MEMORY_POOL}'.\n\n"
+        "The R `arrow` package and `pyarrow` each bundle their own build of "
+        "the libarrow C++ library. Loaded together, they must use the system "
+        "allocator to share buffers. With any other allocator, data sent "
+        "between R and Python is silently replaced by zeros.\n\n"
+        f"Unset {ARROW_MEMORY_POOL_ENVVAR}, or set it before starting Python:\n\n"
+        f"    export {ARROW_MEMORY_POOL_ENVVAR}={ARROW_MEMORY_POOL}"
+    )
+
+
+def arrow_interop_error(observed: list) -> RuntimeError:
+    """Build the error raised when an Arrow round-trip loses data.
+
+    Args:
+        observed (list): Values read back from R for `PROBE_VALUES`.
+
+    Returns:
+        RuntimeError: Error describing the cause and how to resolve it.
+    """
+    import pyarrow as pa
+
+    pool = pa.default_memory_pool().backend_name
+
+    return RuntimeError(
+        "Data sent to R is coming back corrupted, so pysits cannot run.\n\n"
+        f"Sent {PROBE_VALUES}, received {observed}.\n\n"
+        "The R `arrow` package and `pyarrow` each bundle their own build of "
+        "the libarrow C++ library. Loaded together, they must use the system "
+        f"allocator to share buffers, but pyarrow is using '{pool}'.\n\n"
+        "This happens when pyarrow is initialized before pysits with a "
+        "different allocator. Either import pysits before pyarrow, or set the "
+        "environment variable before starting Python:\n\n"
+        f"    export {ARROW_MEMORY_POOL_ENVVAR}={ARROW_MEMORY_POOL}"
+    )

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/pysits/backend/pkgs.py

@@ -17,6 +17,7 @@
 
 """backend packages."""
 
+from pysits.backend.arrow import check_arrow_memory_pool
 from pysits.backend.loaders import load_package
 from pysits.settings import __sitsver__
 
@@ -33,4 +34,9 @@ r_pkg_leaflet = load_package("leaflet")
 r_pkg_kohonen = load_package("kohonen")
 r_pkg_sf = load_package("sf")
 r_pkg_htmlwidgets = load_package("htmlwidgets")
+
+# `arrow` brings a second libarrow build into the process, so the allocator has
+# to be compatible before it is loaded. See `pysits.backend.arrow`.
+check_arrow_memory_pool()
+
 r_pkg_arrow = load_package("arrow")

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/pysits/conversions/tibble.py

@@ -36,6 +36,7 @@ from rpy2.robjects.conversion import localconverter
 from rpy2.robjects.vectors import DataFrame as RDataFrame
 from shapely import wkt
 
+from pysits.backend.arrow import PROBE_COLUMN, PROBE_VALUES, arrow_interop_error
 from pysits.backend.functions import r_fnc_class, r_fnc_set_column
 from pysits.backend.pkgs import r_pkg_base, r_pkg_sf, r_pkg_sits
 from pysits.models.frame import NestedFrame
@@ -90,6 +91,39 @@ def _sf_to_shapely(sf_object: RDataFrame) -> list:
 #
 # Arrow IPC helpers
 #
+def _check_arrow_interop() -> None:
+    """Verify that Arrow data is not corrupted once R's libarrow is loaded.
+
+    R libarrow only initializes on first use, so a probe is sent through R to
+    force it, and ``pyarrow`` is then re-tested. A mismatched pair of libarrow
+    builds corrupts the conversion of Python objects to Arrow arrays, returning
+    zeros instead of raising, which would silently damage every nested column.
+
+    Raises:
+        RuntimeError: If either transfer loses data.
+    """
+    probe = PandasDataFrame({PROBE_COLUMN: PROBE_VALUES})
+
+    # Forces R libarrow to initialize, and checks the Python -> R direction
+    r_probe = rpy2_globalenv["pysits_read_ipc_raw"](
+        robjects.vectors.ByteVector(_dataframe_to_ipc_bytes(probe))
+    )
+
+    # Load content from R
+    observed = list(r_probe.rx2(PROBE_COLUMN))
+
+    # If it is different, then given an error
+    if observed != PROBE_VALUES:
+        raise arrow_interop_error(observed)
+
+    # Both libarrow builds are live now: check the conversion of Python objects
+    # to Arrow arrays, which is the path nested columns depend on.
+    observed = pa.array(PROBE_VALUES).to_pylist()
+
+    if observed != PROBE_VALUES:
+        raise arrow_interop_error(observed)
+
+
 def _ensure_r_ipc_functions():
     """Define R-side IPC reader/writer/unnester functions once."""
     if "pysits_write_ipc_raw" in rpy2_globalenv:
@@ -162,6 +196,8 @@ def _ensure_r_ipc_functions():
         }
     """)
 
+    _check_arrow_interop()
+
 
 def _dataframe_to_ipc_bytes(df: PandasDataFrame) -> bytes:
     """Serialize a pandas DataFrame to Arrow IPC bytes.
@@ -480,7 +516,8 @@ def geopandas_to_tibble(data: GeoPandasDataFrame) -> RDataFrame:
             f"Warning: Dropping columns with embedded DataFrames: {dropped_columns}"
         )
 
-    data_safe = data[safe_columns].copy()
+    # Geometries are transferred as WKT, so the result is a plain data frame
+    data_safe = PandasDataFrame(data[safe_columns].copy())
 
     if isinstance(data, GeoPandasDataFrame):
         geom_col = data.geometry.name

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/pysits/docs/content/impute_linear.md

@@ -3,7 +3,8 @@ Replace NA values by linear interpolation
 Remove NA by linear interpolation
 
 Args:
-    data (list | pandas.DataFrame): A time series vector or matrix.
+    data (list): A time series vector or matrix.
 
 Returns:
     R: A set of filtered time series using the imputation function.
+

pysits-2.0.0.dev1/pysits/docs/content/impute_median.md → pysits-2.0.0.dev3/pysits/docs/content/impute_mean.md

@@ -1,6 +1,6 @@
-Remove NA using median
+Remove NA using mean
 
-Remove NA using median
+Remove NA using mean
 
 Args:
     data (list[float] | SITSMatrix): A time series vector or matrix.

{pysits-2.0.0.dev1 → pysits-2.0.0.dev3}/pysits/docs/content/impute_mean_window.md

@@ -3,14 +3,14 @@ Remove NA using weighted moving average
 Remove NA using weighted moving average
 
 Args:
-    data (list): A time series vector or matrix.
+    data (list[float]): A time series vector or matrix.
     k (int): Width of the moving average window. Expands to both sides
         of the center element e.g. k = 2 means 4 observations (2 left,
         2 right) are taken into account. If all observations in the
         current window are NA, the window size is automatically
         increased until there are at least 2 non-NA values present.
-    weighting (str): The weighting strategy to be used. More details
-        below (default is "simple").
+    weighting (str): Weighting strategy to be used. More details below
+        (default is "simple").
 
 Returns:
     R: A set of filtered time series using the imputation function.

pysits-2.0.0.dev3/pysits/docs/content/impute_median.md

@@ -0,0 +1,9 @@
+Remove NA using median
+
+Remove NA using median
+
+Args:
+    data (list | SITSMatrix): A time series vector or matrix.
+
+Returns:
+    R: A set of filtered time series using the imputation function.