Refactor bf2raw command to be centrally placed

.github/workflows/main.yml

@@ -104,6 +104,10 @@ jobs:
           python -m pip install --upgrade pip
           pip install -e . -r requirements.txt
 
+      - name: Directory setup
+        run: |
+          cp .env.sample .env
+
       - name: Test Coverage
         continue-on-error: true
         run: |

.gitignore

@@ -52,6 +52,8 @@ coverage.xml
 .hypothesis/
 .pytest_cache/
 
+test/tmp/
+
 # Translations
 *.mo
 *.pot

em_workflows/brt/flow.py

@@ -35,6 +35,7 @@
 - Finally, we ``POST`` the JSON datastructure to the API, and cleanup temp dirs.
 """
 
+from typing import Dict
 import glob
 import os
 from em_workflows.file_path import FilePath
@@ -45,6 +46,7 @@
 from prefect import task, Flow, Parameter, unmapped
 from prefect.run_configs import LocalRun
 from prefect.engine import signals
+from pytools.workflow_functions import visual_min_max
 
 from em_workflows.utils import utils
 from em_workflows.utils import neuroglancer as ng
@@ -475,6 +477,41 @@ def cleanup_files(file_path: FilePath, pattern=str):
 #    return inputs_paired
 
 
+@task
+def gen_zarr(fp_in: FilePath):
+    file_path = fp_in
+    # fallback mrc file
+    input_file = file_path.fp_in.as_posix()
+
+    rec_mrc = file_path.gen_output_fp(output_ext="_rec.mrc")
+    if rec_mrc.is_file():
+        input_file = rec_mrc.as_posix()
+
+    ng.bioformats_gen_zarr(
+        file_path=file_path,
+        input_fname=input_file,
+        depth=BRT_DEPTH,
+        width=BRT_WIDTH,
+        height=BRT_HEIGHT,
+        resolutions=1,
+    )
+    ng.zarr_build_multiscales(file_path)
+
+
+@task
+def gen_ng_metadata(fp_in: FilePath) -> Dict:
+    file_path = fp_in
+    asset_fp = Path(f"{file_path.assets_dir}/{file_path.base}.zarr")
+
+    first_zarr_arr = Path(asset_fp.as_posix() + "/0")
+
+    ng_asset = file_path.gen_asset(
+        asset_type=AssetType.NEUROGLANCER_ZARR, asset_fp=asset_fp
+    )
+    ng_asset["metadata"] = visual_min_max(mad_scale=5, input_image=first_zarr_arr)
+    return ng_asset
+
+
 with Flow(
     "brt_flow",
     executor=BRTConfig.SLURM_EXECUTOR,
@@ -595,14 +632,8 @@ def cleanup_files(file_path: FilePath, pattern=str):
     )
     # finished volslicer inputs.
 
-    # START PYRAMID GEN
-    pyramid_assets = ng.gen_zarr.map(
-        fp_in=fps,
-        depth=unmapped(BRT_DEPTH),
-        width=unmapped(BRT_WIDTH),
-        height=unmapped(BRT_HEIGHT),
-        upstream_tasks=[brts],
-    )
+    zarrs = gen_zarr.map(fp_in=fps, upstream_tasks=[brts])
+    pyramid_assets = gen_ng_metadata.map(fp_in=fps, upstream_tasks=[zarrs])
     #  archive_pyramid_cmds = ng.gen_archive_pyr.map(
     #      file_path=fps, upstream_tasks=[pyramid_assets]
     #  )

em_workflows/constants.py

@@ -2,6 +2,7 @@
 
 LARGE_DIM = 1024
 SMALL_DIM = 300
+RECHUNK_SIZE = 512
 
 BIOFORMATS_NUM_WORKERS = 3
 # This is expected to be less than the available memory for a dask worker

em_workflows/czi/constants.py

@@ -1,5 +1,4 @@
 VALID_CZI_INPUTS = ["czi", "CZI"]
-RECHUNK_SIZE = 512
 SITK_COMPRESSION_LVL = 90
 THUMB_X_DIM = 300
 THUMB_Y_DUM = 300

em_workflows/czi/flow.py

@@ -6,10 +6,9 @@
 from pytools.HedwigZarrImages import HedwigZarrImages
 from em_workflows.file_path import FilePath
 from em_workflows.utils import utils
+from em_workflows.utils import neuroglancer as ng
 from prefect.run_configs import LocalRun
-from em_workflows.constants import BIOFORMATS_NUM_WORKERS
 from em_workflows.czi.constants import (
-    RECHUNK_SIZE,
     VALID_CZI_INPUTS,
     THUMB_X_DIM,
     THUMB_Y_DUM,
@@ -18,12 +17,6 @@
 from em_workflows.czi.config import CZIConfig
 
 
-def rechunk_zarr(zarr_fp: Path) -> None:
-    images = HedwigZarrImages(zarr_fp, read_only=False)
-    for _, image in images.series():
-        image.rechunk(RECHUNK_SIZE)
-
-
 def gen_thumb(image: HedwigZarrImage, file_path: FilePath, image_name: str) -> dict:
     sitk_image = image.extract_2d(
         target_size_x=THUMB_X_DIM, target_size_y=THUMB_Y_DUM, auto_uint8=True
@@ -84,42 +77,14 @@ def gen_imageSet(file_path: FilePath) -> List:
     return image_set
 
 
-def bioformats_gen_zarr(file_path: FilePath):
-    """
-    TODO, refactor this into ng.gen_zarr
-    bioformats2raw --max_workers=$nproc  --downsample-type AREA
-    --compression=blosc --compression-properties cname=zstd
-    --compression-properties clevel=5 --compression-properties shuffle=1
-    input.tiff output.zarr
-    """
-    input_czi = f"{file_path.proj_dir}/{file_path.base}.czi"
-    output_zarr = f"{file_path.working_dir}/{file_path.base}.zarr"
-    log_fp = f"{file_path.working_dir}/{file_path.base}_as_zarr.log"
-    cmd = [
-        CZIConfig.bioformats2raw,
-        f"--max_workers={BIOFORMATS_NUM_WORKERS}",
-        "--overwrite",
-        "--downsample-type",
-        "AREA",
-        "--compression=blosc",
-        "--compression-properties",
-        "cname=zstd",
-        "--compression-properties",
-        "clevel=5",
-        "--compression-properties",
-        "shuffle=1",
-        input_czi,
-        output_zarr,
-    ]
-    FilePath.run(cmd, log_fp)
-    output_zarr = f"{file_path.working_dir}/{file_path.base}.zarr"
-    file_path.copy_to_assets_dir(fp_to_cp=Path(output_zarr))
-    rechunk_zarr(zarr_fp=Path(output_zarr))
-
-
 @task
 def generate_czi_imageset(file_path: FilePath):
-    bioformats_gen_zarr(file_path)
+    input_czi = f"{file_path.proj_dir}/{file_path.base}.czi"
+    ng.bioformats_gen_zarr(
+        file_path=file_path,
+        input_fname=input_czi,
+        rechunk=True,
+    )
     imageSet = gen_imageSet(file_path=file_path)
     # extract images from input file, used to create imageSet elements
     return imageSet

em_workflows/lrg_2d_rgb/flow.py

@@ -1,13 +1,14 @@
-import os
+from typing import Dict
 from pathlib import Path
 import SimpleITK as sitk
 from pytools import HedwigZarrImage, HedwigZarrImages
 from prefect import Flow, task, Parameter
 from prefect.run_configs import LocalRun
 
 from em_workflows.utils import utils
+from em_workflows.utils import neuroglancer as ng
 from em_workflows.file_path import FilePath
-from em_workflows.constants import AssetType, BIOFORMATS_NUM_WORKERS, JAVA_MAX_HEAP_SIZE
+from em_workflows.constants import AssetType
 from em_workflows.lrg_2d_rgb.config import LRG2DConfig
 from em_workflows.lrg_2d_rgb.constants import (
     LARGE_THUMB_X,
@@ -46,45 +47,23 @@ def convert_png_to_tiff(file_path: FilePath):
 
 
 @task
-def bioformats_gen_zarr(file_path: FilePath):
-    """
-    TODO, refactor this into ng.gen_zarr
-    bioformats2raw --scale-format-string '%2$d' --downsample-type AREA
-    --compression=blosc --compression-properties cname=zstd
-    --compression-properties clevel=5 --compression-properties shuffle=1
-    input.tiff output.zarr
-    """
-
+def gen_zarr(file_path: FilePath) -> None:
     input_tiff = f"{file_path.working_dir}/{file_path.base}.tiff"
-    output_zarr = f"{file_path.working_dir}/{file_path.base}.zarr"
-    log_fp = f"{file_path.working_dir}/{file_path.base}_as_zarr.log"
-    # increase java max heap size for bf2raw command (just in case)
-    os.environ["_JAVA_OPTIONS"] = JAVA_MAX_HEAP_SIZE
-    cmd = [
-        LRG2DConfig.bioformats2raw,
-        f"--max_workers={BIOFORMATS_NUM_WORKERS}",
-        "--overwrite",
-        "--scale-format-string",
-        "%2$d",
-        "--downsample-type",
-        "AREA",
-        "--compression=blosc",
-        "--compression-properties",
-        "cname=zstd",
-        "--compression-properties",
-        "clevel=5",
-        "--compression-properties",
-        "shuffle=1",
-        input_tiff,
-        output_zarr,
-    ]
-    FilePath.run(cmd, log_fp)
-    zarr_images = HedwigZarrImages(zarr_path=Path(output_zarr), read_only=False)
-    zarr_image: HedwigZarrImage = zarr_images[list(zarr_images.get_series_keys())[0]]
-    zarr_image.rechunk(512)
-    asset_fp = file_path.copy_to_assets_dir(fp_to_cp=Path(output_zarr))
+
+    ng.bioformats_gen_zarr(
+        file_path=file_path,
+        input_fname=input_tiff,
+        rechunk=True,
+    )
+
+
+@task
+def generate_ng_asset(file_path: FilePath) -> Dict:
+    output_zarr = Path(f"{file_path.assets_dir}/{file_path.base}.zarr")
+    zarr_images = HedwigZarrImages(zarr_path=output_zarr, read_only=False)
+    zarr_image = zarr_images[list(zarr_images.get_series_keys())[0]]
     ng_asset = file_path.gen_asset(
-        asset_type=AssetType.NEUROGLANCER_ZARR, asset_fp=asset_fp
+        asset_type=AssetType.NEUROGLANCER_ZARR, asset_fp=output_zarr
     )
     ng_asset["metadata"] = dict(
         shader=zarr_image.shader_type,
@@ -160,7 +139,8 @@ def gen_thumb(file_path: FilePath):
     )
     fps = utils.gen_fps(share_name=file_share, input_dir=input_dir_fp, fps_in=input_fps)
     tiffs = convert_png_to_tiff.map(file_path=fps)
-    zarr_assets = bioformats_gen_zarr.map(file_path=fps, upstream_tasks=[tiffs])
+    zarrs = gen_zarr.map(file_path=fps, upstream_tasks=[tiffs])
+    zarr_assets = generate_ng_asset.map(file_path=fps, upstream_tasks=[zarrs])
     thumb_assets = gen_thumb.map(file_path=fps, upstream_tasks=[zarr_assets])
     prim_fps = utils.gen_prim_fps.map(fp_in=fps)
     callback_with_thumbs = utils.add_asset.map(prim_fp=prim_fps, asset=thumb_assets)

em_workflows/sem_tomo/flow.py

@@ -1,13 +1,15 @@
-from em_workflows.file_path import FilePath
+from pathlib import Path
 import glob
 from natsort import os_sorted
 import math
 from typing import Dict
 from prefect import Flow, task, Parameter, unmapped
 from prefect.run_configs import LocalRun
+from pytools.workflow_functions import visual_min_max
 
 from em_workflows.utils import utils
 from em_workflows.utils import neuroglancer as ng
+from em_workflows.file_path import FilePath
 from em_workflows.constants import AssetType
 from em_workflows.sem_tomo.config import SEMConfig
 from em_workflows.sem_tomo.constants import FIBSEM_DEPTH, FIBSEM_HEIGHT, FIBSEM_WIDTH
@@ -207,6 +209,41 @@ def gen_keyimg_small(fp_in: FilePath) -> Dict:
     return keyimg_asset
 
 
+@task
+def gen_zarr(fp_in: FilePath) -> None:
+    file_path = fp_in
+    # fallback mrc file
+    input_file = file_path.fp_in.as_posix()
+
+    base_mrc = file_path.gen_output_fp(output_ext=".mrc", out_fname="adjusted.mrc")
+    if base_mrc.is_file():
+        input_file = base_mrc.as_posix()
+
+    ng.bioformats_gen_zarr(
+        file_path=file_path,
+        input_fname=input_file,
+        depth=FIBSEM_DEPTH,
+        width=FIBSEM_WIDTH,
+        height=FIBSEM_HEIGHT,
+        resolutions=1,
+    )
+    ng.zarr_build_multiscales(fp_in)
+
+
+@task
+def gen_ng_metadata(fp_in: FilePath) -> Dict:
+    file_path = fp_in
+    asset_fp = Path(f"{file_path.assets_dir}/{file_path.base}.zarr")
+
+    first_zarr_arr = Path(asset_fp.as_posix() + "/0")
+
+    ng_asset = file_path.gen_asset(
+        asset_type=AssetType.NEUROGLANCER_ZARR, asset_fp=asset_fp
+    )
+    ng_asset["metadata"] = visual_min_max(mad_scale=5, input_image=first_zarr_arr)
+    return ng_asset
+
+
 with Flow(
     "sem_tomo",
     state_handlers=[utils.notify_api_completion, utils.notify_api_running],
@@ -278,13 +315,8 @@ def gen_keyimg_small(fp_in: FilePath) -> Dict:
     thumb_assets = gen_keyimg_small.map(fp_in=fps, upstream_tasks=[keyimg_assets])
 
     # zarr file generation
-    pyramid_assets = ng.gen_zarr.map(
-        fp_in=fps,
-        depth=unmapped(FIBSEM_DEPTH),
-        width=unmapped(FIBSEM_WIDTH),
-        height=unmapped(FIBSEM_HEIGHT),
-        upstream_tasks=[base_mrcs],
-    )
+    zarrs = gen_zarr.map(fp_in=fps, upstream_tasks=[base_mrcs])
+    pyramid_assets = gen_ng_metadata.map(fp_in=fps, upstream_tasks=[zarrs])
 
     # this is the toplevel element (the input file basically) onto which
     # the "assets" (ie the outputs derived from this file) are hung.