rfmix-reader 0.2.0__tar.gz → 0.2.1__tar.gz

rfmix_reader-0.2.1/PKG-INFO

@@ -0,0 +1,259 @@
+Metadata-Version: 2.4
+Name: rfmix-reader
+Version: 0.2.1
+Summary: RFMix-reader is a Python package designed to efficiently read and process output files generated by RFMix, a popular tool for estimating local ancestry in admixed populations. The package employs a lazy loading approach, which minimizes memory consumption by reading only the loci that are accessed by the user, rather than loading the entire dataset into memory at once.
+License: GPL-3.0-or-later
+License-File: LICENSE
+Keywords: file parser,rfmix,gpu acceleration,local ancestry
+Author: Kynon J.M. Benjamin
+Author-email: kj.benjamin90@gmail.com
+Maintainer: Kynon J.M. Benjamin
+Maintainer-email: kj.benjamin90@gmail.com
+Requires-Python: >=3.11,<3.15
+Classifier: License :: OSI Approved :: GNU General Public License v3 or later (GPLv3+)
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Provides-Extra: docs
+Provides-Extra: gpu
+Provides-Extra: io
+Provides-Extra: tests
+Provides-Extra: viz
+Requires-Dist: cairosvg (>=2.7,<3.0) ; extra == "viz"
+Requires-Dist: cyvcf2 (>=0.31)
+Requires-Dist: dask (>=2025.1,<2026.0)
+Requires-Dist: matplotlib (>=3.10,<4.0) ; extra == "viz"
+Requires-Dist: numpy (>=1.23,<3)
+Requires-Dist: pandas (>=2.0)
+Requires-Dist: psutil (>=6,<7)
+Requires-Dist: seaborn (>=0.13,<0.14) ; extra == "viz"
+Requires-Dist: sphinx (>=7,<9) ; extra == "docs"
+Requires-Dist: sphinx-autodoc-typehints (>=2,<3) ; extra == "docs"
+Requires-Dist: sphinx-copybutton (>=0.5,<0.6) ; extra == "docs"
+Requires-Dist: sphinx-rtd-theme (>=2,<3) ; extra == "docs"
+Requires-Dist: torch (>=2.8) ; extra == "gpu"
+Requires-Dist: tqdm (>=4.66)
+Project-URL: Bug Tracker, https://github.com/heart-gen/rfmix_reader/issues
+Project-URL: homepage, https://rfmix-reader.readthedocs.io/en/latest/
+Project-URL: repository, https://github.com/heart-gen/rfmix_reader.git
+Description-Content-Type: text/markdown
+
+# RFMix-reader
+`RFMix-reader` is a Python package for efficiently reading and processing output
+files generated by [`RFMix`](https://github.com/slowkoni/rfmix), a widely used tool
+for estimating local ancestry in admixed populations.  
+It employs a **lazy loading approach** to minimize memory usage, and leverages **GPU acceleration**
+for major speedups when available.
+
+---
+
+## Installation
+
+`rfmix-reader` requires **Python 3.10+**. Install from PyPI:
+
+```bash
+pip install rfmix-reader
+````
+
+### Installation Options
+
+* **Basic install** (CPU only):
+
+  ```bash
+  pip install rfmix-reader
+  ```
+
+* **With GPU acceleration** (`cupy`, `cudf`, `dask-cudf`):
+
+  ```bash
+  pip install rfmix-reader[gpu]
+  ```
+
+* **With documentation tools** (`sphinx`, `sphinx-rtd-theme`):
+
+  ```bash
+  pip install rfmix-reader[docs]
+  ```
+
+* **With testing tools** (`pytest`):
+
+  ```bash
+  pip install rfmix-reader[tests]
+  ```
+
+### GPU Notes
+
+* `torch` is installed automatically.
+* For CUDA builds, install a matching GPU-enabled wheel for your system following the [PyTorch guide](https://pytorch.org/get-started/locally/).
+* RAPIDS (`cudf`, `cupy`) wheels are version- and CUDA-specific. See the [RAPIDS install guide](https://docs.rapids.ai/install).
+* CPU-only installations will still run efficiently, just without GPU acceleration.
+
+---
+
+## Quickstart
+
+```python
+from rfmix_reader import read_rfmix
+
+# Load RFMix outputs (two-population admixture example)
+file_path = "examples/two_populations/out/"
+loci_df, g_anc, local_array = read_rfmix(file_path)
+
+print(loci_df.head())
+print(g_anc.head())
+print(local_array.shape)
+```
+
+---
+
+## Key Features
+
+* **Lazy Loading**: Reads data on-the-fly, reducing memory footprint.
+* **Efficient Access**: Query specific loci or regions of interest.
+* **Seamless Integration**: Works smoothly with `pandas`, `dask`, and other analysis tools.
+* **Loci Imputation**: Impute local ancestry loci to dense genotype variant sites.
+* **GPU Acceleration**: Automatic CUDA acceleration via PyTorch/CuPy when available.
+
+---
+
+## Simulation Data
+
+Test datasets for two- and three-population admixture are available on Synapse:
+[Synapse Project syn61691659](https://www.synapse.org/Synapse:syn61691659).
+
+---
+
+## Usage
+
+### Binary Conversion
+
+RFMix does not generate binary files directly.
+Use `create_binaries` to generate them (also available as a CLI):
+
+```bash
+create-binaries two_pops/out/
+```
+
+```python
+from rfmix_reader import create_binaries
+
+create_binaries("two_pops/out/", binary_dir="./binary_files")
+```
+
+### Main Function
+
+Once binaries are available, process RFMix results:
+
+```python
+from rfmix_reader import read_rfmix
+
+loci, g_anc, admix = read_rfmix("two_pops/out/")
+```
+
+### Three Population Example
+
+Binaries can also be generated on-the-fly within `read_rfmix` with
+`generate_binary` set to `True`.
+
+```python
+loci, g_anc, admix = read_rfmix("examples/three_populations/out/",
+                               binary_dir="./binary_files",
+                               generate_binary=True)
+```
+
+### Loci Imputation
+
+Impute local ancestry loci to variant positions for integration with genotype data:
+
+```python
+from rfmix_reader import interpolate_array
+import pandas as pd
+import dask.array as da
+
+variant_loci_df = pd.DataFrame({
+    "chrom": ["1", "1", "1", "1"],
+    "pos": [100, 200, 300, 400],
+    "i": [1, None, None, 2]
+})
+admix = da.random.random((2, 3))  # mock admixture data
+
+z = interpolate_array(variant_loci_df, admix, "/path/to/output")
+print(z.shape)
+```
+
+### Reading Haptools simulations
+
+Use `read_simu` to load BGZF-compressed VCF files created by
+`haptools simgenotype --pop_field`:
+
+```python
+from rfmix_reader import read_simu
+
+loci_df, g_anc, admix = read_simu("/path/to/simulations/")
+```
+
+Haptools does **not** include the chromosome length in the `##contig`
+header lines, but `read_simu` requires that metadata to index each VCF.
+Copy the `contigs.txt` file Haptools generates from the FASTA you used
+for simulation and reheader every file with the appropriate contig entry
+before calling `read_simu`. The following snippet shows one approach
+using `bcftools` and `tabix`:
+
+```bash
+CONTIGS="../../three_populations/_m/contigs.txt"
+VCFDIR="gt-files"
+CHR="chr${SLURM_ARRAY_TASK_ID}"
+OUT="${VCFDIR}/${CHR}.vcf.gz"
+IN="${VCFDIR}/back/${CHR}.vcf.gz"
+
+CONTIG_LINE=$(grep -w "ID=${CHR}" "$CONTIGS")
+if [[ -z "$CONTIG_LINE" ]]; then
+    echo "ERROR: No contig line found for ${CHR} in $CONTIGS"
+    exit 1
+fi
+
+bcftools view -h "$IN" \
+    | sed "s/^##contig=<ID=${CHR}>.*/${CONTIG_LINE}/" > header.${CHR}.tmp
+bcftools reheader -h header.${CHR}.tmp -o "$OUT" "$IN"
+tabix -p vcf "$OUT"
+```
+
+---
+
+## Development Install
+
+For contributors:
+
+```bash
+git clone https://github.com/heart-gen/rfmix_reader.git
+cd rfmix_reader
+pip install -e ".[gpu,docs,tests]"
+```
+
+---
+
+## Citation
+
+If you use this software, please cite:
+
+[![DOI](https://zenodo.org/badge/807052842.svg)](https://zenodo.org/doi/10.5281/zenodo.12629787)
+
+Benjamin, K. J. M. (2024). **RFMix-reader (Version 0.2.0)** \[Computer software].
+[https://github.com/heart-gen/rfmix\_reader](https://github.com/heart-gen/rfmix_reader)
+
+Kynon JM Benjamin. *"RFMix-reader: Accelerated reading and processing for local ancestry studies."*
+**bioRxiv** (2024).
+DOI: [10.1101/2024.07.13.603370](https://www.biorxiv.org/content/10.1101/2024.07.13.603370v2).
+
+---
+
+## Funding
+
+This work was supported by the National Institutes of Health,
+National Institute on Minority Health and Health Disparities (NIMHD)
+K99MD016964 / R00MD016964.
+
+

rfmix_reader-0.2.1/README.md

@@ -0,0 +1,215 @@
+# RFMix-reader
+`RFMix-reader` is a Python package for efficiently reading and processing output
+files generated by [`RFMix`](https://github.com/slowkoni/rfmix), a widely used tool
+for estimating local ancestry in admixed populations.  
+It employs a **lazy loading approach** to minimize memory usage, and leverages **GPU acceleration**
+for major speedups when available.
+
+---
+
+## Installation
+
+`rfmix-reader` requires **Python 3.10+**. Install from PyPI:
+
+```bash
+pip install rfmix-reader
+````
+
+### Installation Options
+
+* **Basic install** (CPU only):
+
+  ```bash
+  pip install rfmix-reader
+  ```
+
+* **With GPU acceleration** (`cupy`, `cudf`, `dask-cudf`):
+
+  ```bash
+  pip install rfmix-reader[gpu]
+  ```
+
+* **With documentation tools** (`sphinx`, `sphinx-rtd-theme`):
+
+  ```bash
+  pip install rfmix-reader[docs]
+  ```
+
+* **With testing tools** (`pytest`):
+
+  ```bash
+  pip install rfmix-reader[tests]
+  ```
+
+### GPU Notes
+
+* `torch` is installed automatically.
+* For CUDA builds, install a matching GPU-enabled wheel for your system following the [PyTorch guide](https://pytorch.org/get-started/locally/).
+* RAPIDS (`cudf`, `cupy`) wheels are version- and CUDA-specific. See the [RAPIDS install guide](https://docs.rapids.ai/install).
+* CPU-only installations will still run efficiently, just without GPU acceleration.
+
+---
+
+## Quickstart
+
+```python
+from rfmix_reader import read_rfmix
+
+# Load RFMix outputs (two-population admixture example)
+file_path = "examples/two_populations/out/"
+loci_df, g_anc, local_array = read_rfmix(file_path)
+
+print(loci_df.head())
+print(g_anc.head())
+print(local_array.shape)
+```
+
+---
+
+## Key Features
+
+* **Lazy Loading**: Reads data on-the-fly, reducing memory footprint.
+* **Efficient Access**: Query specific loci or regions of interest.
+* **Seamless Integration**: Works smoothly with `pandas`, `dask`, and other analysis tools.
+* **Loci Imputation**: Impute local ancestry loci to dense genotype variant sites.
+* **GPU Acceleration**: Automatic CUDA acceleration via PyTorch/CuPy when available.
+
+---
+
+## Simulation Data
+
+Test datasets for two- and three-population admixture are available on Synapse:
+[Synapse Project syn61691659](https://www.synapse.org/Synapse:syn61691659).
+
+---
+
+## Usage
+
+### Binary Conversion
+
+RFMix does not generate binary files directly.
+Use `create_binaries` to generate them (also available as a CLI):
+
+```bash
+create-binaries two_pops/out/
+```
+
+```python
+from rfmix_reader import create_binaries
+
+create_binaries("two_pops/out/", binary_dir="./binary_files")
+```
+
+### Main Function
+
+Once binaries are available, process RFMix results:
+
+```python
+from rfmix_reader import read_rfmix
+
+loci, g_anc, admix = read_rfmix("two_pops/out/")
+```
+
+### Three Population Example
+
+Binaries can also be generated on-the-fly within `read_rfmix` with
+`generate_binary` set to `True`.
+
+```python
+loci, g_anc, admix = read_rfmix("examples/three_populations/out/",
+                               binary_dir="./binary_files",
+                               generate_binary=True)
+```
+
+### Loci Imputation
+
+Impute local ancestry loci to variant positions for integration with genotype data:
+
+```python
+from rfmix_reader import interpolate_array
+import pandas as pd
+import dask.array as da
+
+variant_loci_df = pd.DataFrame({
+    "chrom": ["1", "1", "1", "1"],
+    "pos": [100, 200, 300, 400],
+    "i": [1, None, None, 2]
+})
+admix = da.random.random((2, 3))  # mock admixture data
+
+z = interpolate_array(variant_loci_df, admix, "/path/to/output")
+print(z.shape)
+```
+
+### Reading Haptools simulations
+
+Use `read_simu` to load BGZF-compressed VCF files created by
+`haptools simgenotype --pop_field`:
+
+```python
+from rfmix_reader import read_simu
+
+loci_df, g_anc, admix = read_simu("/path/to/simulations/")
+```
+
+Haptools does **not** include the chromosome length in the `##contig`
+header lines, but `read_simu` requires that metadata to index each VCF.
+Copy the `contigs.txt` file Haptools generates from the FASTA you used
+for simulation and reheader every file with the appropriate contig entry
+before calling `read_simu`. The following snippet shows one approach
+using `bcftools` and `tabix`:
+
+```bash
+CONTIGS="../../three_populations/_m/contigs.txt"
+VCFDIR="gt-files"
+CHR="chr${SLURM_ARRAY_TASK_ID}"
+OUT="${VCFDIR}/${CHR}.vcf.gz"
+IN="${VCFDIR}/back/${CHR}.vcf.gz"
+
+CONTIG_LINE=$(grep -w "ID=${CHR}" "$CONTIGS")
+if [[ -z "$CONTIG_LINE" ]]; then
+    echo "ERROR: No contig line found for ${CHR} in $CONTIGS"
+    exit 1
+fi
+
+bcftools view -h "$IN" \
+    | sed "s/^##contig=<ID=${CHR}>.*/${CONTIG_LINE}/" > header.${CHR}.tmp
+bcftools reheader -h header.${CHR}.tmp -o "$OUT" "$IN"
+tabix -p vcf "$OUT"
+```
+
+---
+
+## Development Install
+
+For contributors:
+
+```bash
+git clone https://github.com/heart-gen/rfmix_reader.git
+cd rfmix_reader
+pip install -e ".[gpu,docs,tests]"
+```
+
+---
+
+## Citation
+
+If you use this software, please cite:
+
+[![DOI](https://zenodo.org/badge/807052842.svg)](https://zenodo.org/doi/10.5281/zenodo.12629787)
+
+Benjamin, K. J. M. (2024). **RFMix-reader (Version 0.2.0)** \[Computer software].
+[https://github.com/heart-gen/rfmix\_reader](https://github.com/heart-gen/rfmix_reader)
+
+Kynon JM Benjamin. *"RFMix-reader: Accelerated reading and processing for local ancestry studies."*
+**bioRxiv** (2024).
+DOI: [10.1101/2024.07.13.603370](https://www.biorxiv.org/content/10.1101/2024.07.13.603370v2).
+
+---
+
+## Funding
+
+This work was supported by the National Institutes of Health,
+National Institute on Minority Health and Health Disparities (NIMHD)
+K99MD016964 / R00MD016964.
+

{rfmix_reader-0.2.0 → rfmix_reader-0.2.1}/pyproject.toml

@@ -1,14 +1,12 @@
 [tool.poetry]
 name = "rfmix-reader"
-version = "0.2.0"  # This will be overridden if dynamic versioning is enabled
+version = "0.2.1"  # This will be overridden if dynamic versioning is enabled
 description = "RFMix-reader is a Python package designed to efficiently read and process output files generated by RFMix, a popular tool for estimating local ancestry in admixed populations. The package employs a lazy loading approach, which minimizes memory consumption by reading only the loci that are accessed by the user, rather than loading the entire dataset into memory at once."
 authors = ["Kynon J.M. Benjamin <kj.benjamin90@gmail.com>"]
 maintainers = ["Kynon J.M. Benjamin <kj.benjamin90@gmail.com>"]
 readme = "README.md"
 license = "GPL-3.0-or-later"
 keywords = ["file parser", "rfmix", "gpu acceleration", "local ancestry"]
-homepage = "https://rfmix-reader.readthedocs.io/en/latest/"
-repository = "https://github.com/heart-gen/rfmix_reader.git"
 packages = [{ include = "rfmix_reader" }]
 classifiers = [
   "Programming Language :: Python :: 3",
@@ -17,28 +15,49 @@ classifiers = [
 ]
 
 [tool.poetry.dependencies]
-python = ">=3.10"
+python = ">=3.11,<3.15"
+numpy = ">=1.23,<3"
 pandas = ">=2.0"
-dask = ">=2025.1"
 tqdm = ">=4.66"
-numpy = ">=2.0"
-zarr = ">=2.0"
-seaborn = "^0.13.2"
-matplotlib = "^3.10.1"
-cairosvg = "^2.7.1"
+psutil = "^6"
+cyvcf2 = ">=0.31"
+dask = "^2025.1"
+# Plotting
+matplotlib = { version = "^3.10", optional = true }
+seaborn = { version = "^0.13", optional = true }
+cairosvg = { version = "^2.7", optional = true }
+# GPU
+torch = { version = ">=2.8", optional = true }
+# Docs
+sphinx                    = { version = ">=7,<9", optional = true }
+sphinx-rtd-theme          = { version = "^2", optional = true }
+sphinx-autodoc-typehints  = { version = "^2", optional = true }
+sphinx-copybutton         = { version = "^0.5", optional = true }
+
+[tool.poetry.extras]
+viz = ["matplotlib", "seaborn", "cairosvg"]
+gpu = ["cupy", "cudf", "dask-cudf", "torch"]
+io  = ["pyarrow"]  # parquet export
+docs = ["sphinx", "sphinx-rtd-theme", "sphinx-autodoc-typehints", "sphinx-copybutton"]
+tests = ["pytest"]
+
+[tool.poetry.urls]
+homepage = "https://rfmix-reader.readthedocs.io/en/latest/"
+repository = "https://github.com/heart-gen/rfmix_reader.git"
+"Bug Tracker" = "https://github.com/heart-gen/rfmix_reader/issues"
 
 [tool.poetry.scripts]
 create-binaries = "rfmix_reader._cli:main"
 
 [tool.poetry.group.test.dependencies]
-pytest = "^8.2"
+pytest = "^8"
+pytest-cov = "^5"
 
 [tool.poetry.group.docs.dependencies]
-sphinx = "^7.3.7"
-sphinx-rtd-theme = "^2.0.0"
-sphinx-autodoc-typehints = "^2.1.0"
-torch = {version = "^2.6+cpu", source = "pytorch_cpu"}
-torchvision = {version = "^0.21+cpu", source = "pytorch_cpu"}
+sphinx = ">=7,<9"
+sphinx-rtd-theme = "^2"
+sphinx-autodoc-typehints = "^2"
+sphinx-copybutton = "^0.5"
 
 [[tool.poetry.source]]
 name = "pytorch_cpu"
@@ -47,7 +66,9 @@ priority = "explicit"
 
 [tool.poetry-dynamic-versioning]
 enable = false
-write_to = "rfmix_reader/_version.py"
+
+[tool.poetry-dynamic-versioning.metadata]
+file = "rfmix_reader/_version.py"
 
 [build-system]
 requires = ["poetry-core>=2.0.0", "poetry-dynamic-versioning"]

rfmix_reader-0.2.1/rfmix_reader/__init__.py

@@ -0,0 +1,92 @@
+from __future__ import annotations
+
+from importlib.metadata import version as _v, PackageNotFoundError
+from typing import TYPE_CHECKING
+
+try:
+    __version__ = _v("rfmix-reader")  # distribution name
+except PackageNotFoundError:
+    try:
+        from ._version import __version__  # fallback for local builds
+    except Exception:
+        __version__ = "0.0.0"
+
+# Public API
+__all__ = [
+    "Chunk",
+    "read_fb", "read_simu", "read_rfmix", "read_flare",
+    "write_data",
+    "admix_to_bed_individual",
+    "CHROM_SIZES", "COORDINATES",
+    "BinaryFileNotFoundError",
+    "interpolate_array",
+    "get_pops", "get_prefixes", "create_binaries", "get_sample_names",
+    "set_gpu_environment", "delete_files_or_directories",
+    "save_multi_format", "generate_tagore_bed",
+    "plot_global_ancestry", "plot_ancestry_by_chromosome",
+    "plot_local_ancestry_tagore",
+]
+
+# Map public names for lazy loading
+_lazy = {
+    "Chunk": ("._chunk", "Chunk"),
+    "read_fb": ("._fb_read", "read_fb"),
+    "read_simu": ("._read_simu", "read_simu"),
+    "read_rfmix": ("._read_rfmix", "read_rfmix"),
+    "read_flare": ("._read_flare", "read_flare"),
+    "write_data": ("._write_data", "write_data"),
+    "admix_to_bed_individual": ("._loci_bed", "admix_to_bed_individual"),
+    "CHROM_SIZES": ("._constants", "CHROM_SIZES"),
+    "COORDINATES": ("._constants", "COORDINATES"),
+    "BinaryFileNotFoundError": ("._errorhandling", "BinaryFileNotFoundError"),
+    "interpolate_array": ("._imputation", "interpolate_array"),
+    "get_pops": ("._utils", "get_pops"),
+    "get_prefixes": ("._utils", "get_prefixes"),
+    "create_binaries": ("._utils", "create_binaries"),
+    "get_sample_names": ("._utils", "get_sample_names"),
+    "set_gpu_environment": ("._utils", "set_gpu_environment"),
+    "delete_files_or_directories": ("._utils", "delete_files_or_directories"),
+    "save_multi_format": ("._visualization", "save_multi_format"),
+    "generate_tagore_bed": ("._visualization", "generate_tagore_bed"),
+    "plot_global_ancestry": ("._visualization", "plot_global_ancestry"),
+    "plot_ancestry_by_chromosome": ("._visualization", "plot_ancestry_by_chromosome"),
+    "plot_local_ancestry_tagore": ("._tagore", "plot_local_ancestry_tagore"),
+}
+
+def __getattr__(name: str):
+    """Lazy attribute loader to keep import-time light."""
+    if name in _lazy:
+        import importlib
+        mod_name, attr_name = _lazy[name]
+        mod = importlib.import_module(mod_name, __name__)
+        obj = getattr(mod, attr_name)
+        globals()[name] = obj  # cache for future access
+        return obj
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
+
+
+def __dir__():
+    # help() and tab-complete show public API
+    return sorted(list(globals().keys()) + __all__)
+
+# Make type checkers happy without importing heavy deps at runtime
+if TYPE_CHECKING:
+    from ._chunk import Chunk
+    from ._fb_read import read_fb
+    from ._read_simu import read_simu
+    from ._read_rfmix import read_rfmix
+    from ._read_flare import read_flare
+    from ._write_data import write_data
+    from ._loci_bed import admix_to_bed_individual
+    from ._constants import CHROM_SIZES, COORDINATES
+    from ._errorhandling import BinaryFileNotFoundError
+    from ._imputation import interpolate_array
+    from ._utils import (
+        get_pops, get_prefixes, create_binaries, get_sample_names,
+        set_gpu_environment, delete_files_or_directories,
+    )
+    from ._visualization import (
+        save_multi_format, generate_tagore_bed,
+        plot_global_ancestry, plot_ancestry_by_chromosome,
+    )
+    from ._tagore import plot_local_ancestry_tagore

{rfmix_reader-0.2.0 → rfmix_reader-0.2.1}/rfmix_reader/_fb_read.py

@@ -5,7 +5,6 @@ Source: https://github.com/limix/pandas-plink/blob/main/pandas_plink/_bed_read.p
 from dask.delayed import delayed
 from dask.array import from_delayed, Array, concatenate
 from numpy import (
-    ascontiguousarray,
     float32,
     memmap,
     int32
@@ -66,7 +65,7 @@ def read_fb(
                 col_end,
             )
             shape = (row_end - row_start, col_end - col_start)
-            row_sx.append(from_delayed(x, shape, dtype=float32))
+            row_sx.append(from_delayed(x, shape, dtype=int32))
             col_start = col_end
         col_sx.append(concatenate(row_sx, 1, True))
         row_start = row_end
@@ -103,5 +102,5 @@ def _read_chunk(
     
     buff = memmap(filepath, dtype=float32, mode="r",
                   offset=offset, shape=size)
-    return ascontiguousarray(buff, int32)
+    return buff.astype(int32, copy=False)