plexora 0.0.1__py3-none-any.whl

plexora/server/utils/native_dialog.py

@@ -0,0 +1,168 @@
+# Native OS file/folder picker for the "Browse..." buttons in the quick-view
+# and import UIs. Returns a path only -- never file bytes.
+#
+# macOS goes through `osascript`'s `choose file`/`choose folder` (a real
+# Cocoa panel driven by AppleScript, no Python GUI toolkit involved) rather
+# than tkinter. tkinter's askopenfilename on macOS routes through a hidden Tk
+# root window that Cocoa uses to size/position/parent the panel, and that
+# path is a long-standing source of broken behavior on recent macOS -- the
+# panel can appear tiny and pinned to a screen corner, barely resizable, or
+# even dismiss itself immediately (the click that opened it double-counts as
+# a click "outside" the misplaced panel). osascript's panel is a normal,
+# independently-managed system dialog and doesn't have any of that.
+#
+# Other platforms keep the tkinter subprocess approach, run out-of-process
+# for two reasons: Tk needs to run on the process's *main* thread on macOS
+# (moot here since macOS no longer uses this path, but still true in
+# principle), but more importantly a request handled by waitress lands on an
+# arbitrary worker thread; and a subprocess crash/hang (e.g. no display on a
+# headless box) can't take the server down with it.
+#
+# Only meant for the terminal/`python run.py` and Jupyter (`server_cli.py`)
+# launch paths, where sys.executable is a real Python interpreter. It is not
+# safe to use as-is from a frozen/packaged build, where sys.executable is the
+# packaged exe itself.
+
+import json
+import subprocess
+import sys
+
+# Keyed by the caller's `file_filter` argument (kept in sync with
+# browsePicker.js's data-browse-filter values).
+_TK_FILTERS = {
+    "image": [("Image files", "*.tif *.tiff *.ome.tif *.ome.tiff *.svs *.qptiff *.png *.jpg *.jpeg"), ("All files", "*.*")],
+    "csv": [("CSV files", "*.csv"), ("All files", "*.*")],
+    "h5ad": [("AnnData files", "*.h5ad"), ("All files", "*.*")],
+    # The single Data input accepts any feature-table format, so its picker
+    # offers all of them at once rather than making the user pick a format
+    # before picking a file. A .zarr store is a directory and is reached by
+    # the same field's "Store..." button in directory mode.
+    "data": [("Single-cell data", "*.csv *.tsv *.txt *.h5ad"), ("All files", "*.*")],
+    "any": [("All files", "*.*")],
+}
+
+#: The filter names callers may ask for. Exported so the route validates
+#: against this table rather than its own hand-typed copy -- the two drifted
+#: once already: "data" was added here for the single Data input on the import
+#: page but not to the route's allowlist, so that field's "File..." button
+#: posted a filter the server rejected with a 400 and did nothing at all.
+FILTER_NAMES = frozenset(_TK_FILTERS)
+
+# AppleScript's `choose file of type` filters by Uniform Type Identifier
+# conformance, not literally by filename extension -- an extension string
+# only works here when macOS actually has a UTI registered for it (true for
+# common ones like these image formats, via system/Preview/etc.
+# declarations). ".h5ad" has no such registration on a stock Mac, so filtering
+# by it doesn't just fail to narrow the list -- it greys out every file,
+# including real .h5ad ones, since none of them "conform" to an unregistered
+# type. Left unfiltered (None -> no `of type` clause) rather than chasing a
+# UTI that may not exist on the user's machine.
+_APPLESCRIPT_EXTENSIONS = {
+    "image": ["tif", "tiff", "svs", "qptiff", "png", "jpg", "jpeg"],
+    "csv": ["csv"],
+    "h5ad": None,
+    # Unfiltered for the same reason as "h5ad" above: the set includes .h5ad,
+    # and one unregistered extension greys out every file in the dialog rather
+    # than just failing to match its own.
+    "data": None,
+    "any": None,
+}
+
+# The dialog script only ever receives a fixed, known filter key on argv --
+# never an interpolated filetypes string -- since the script itself is
+# passed to `python -c`.
+_TK_DIALOG_SCRIPT = r"""
+import json, sys
+import tkinter as tk
+from tkinter import filedialog
+
+mode = sys.argv[1]
+filter_key = sys.argv[2] if len(sys.argv) > 2 else "any"
+
+FILTERS = %(filters)s
+
+root = tk.Tk()
+root.withdraw()
+root.attributes("-topmost", True)
+if mode == "directory":
+    path = filedialog.askdirectory(parent=root)
+else:
+    path = filedialog.askopenfilename(parent=root, filetypes=FILTERS.get(filter_key, FILTERS["any"]))
+root.destroy()
+print(json.dumps({"path": path or None}))
+""" % {"filters": repr(_TK_FILTERS)}
+
+
+def _browse_for_path_macos(mode, file_filter, timeout):
+    if mode == "directory":
+        script = 'POSIX path of (choose folder with prompt "Select a folder")'
+    else:
+        extensions = _APPLESCRIPT_EXTENSIONS.get(file_filter)
+        of_type = ""
+        if extensions:
+            quoted = ", ".join(f'"{ext}"' for ext in extensions)
+            of_type = f" of type {{{quoted}}}"
+        script = f'POSIX path of (choose file{of_type} with prompt "Select a file")'
+
+    try:
+        result = subprocess.run(
+            ["osascript", "-e", script],
+            capture_output=True,
+            text=True,
+            timeout=timeout,
+        )
+    except subprocess.TimeoutExpired:
+        raise RuntimeError("Timed out waiting for the file browser.")
+    except OSError as exc:
+        raise RuntimeError(str(exc))
+
+    if result.returncode != 0:
+        stderr = (result.stderr or "").strip()
+        # AppleScript error -128 is exactly "user cancelled" -- not a failure.
+        if "-128" in stderr or "User canceled" in stderr:
+            return None
+        message = stderr.splitlines()[-1] if stderr else "Could not open the file browser."
+        raise RuntimeError(message)
+
+    path = result.stdout.strip()
+    return path or None
+
+
+def _browse_for_path_tk(mode, file_filter, timeout):
+    try:
+        result = subprocess.run(
+            [sys.executable, "-c", _TK_DIALOG_SCRIPT, mode, file_filter],
+            capture_output=True,
+            text=True,
+            timeout=timeout,
+        )
+    except subprocess.TimeoutExpired:
+        raise RuntimeError("Timed out waiting for the file browser.")
+    except OSError as exc:
+        raise RuntimeError(str(exc))
+
+    if result.returncode != 0:
+        message = (result.stderr or "").strip().splitlines()
+        raise RuntimeError(message[-1] if message else "Could not open the file browser.")
+
+    try:
+        return json.loads(result.stdout.strip().splitlines()[-1])["path"]
+    except (ValueError, IndexError, KeyError):
+        raise RuntimeError("Unexpected response from the file browser.")
+
+
+def browse_for_path(mode="file", file_filter="any", timeout=300):
+    """Opens a native file/folder picker and returns the chosen absolute
+    path, or None if the user cancelled.
+
+    `file_filter` narrows the file-type dropdown for mode="file" (ignored
+    for mode="directory") -- one of "image", "csv", "h5ad", or "any"
+    (default, no narrowing beyond "All files").
+
+    Raises RuntimeError with a user-facing message if no picker could be
+    shown at all (no display, tkinter not installed, timed out, ...) so
+    callers can fall back to a manual path input.
+    """
+    if sys.platform == "darwin":
+        return _browse_for_path_macos(mode, file_filter, timeout)
+    return _browse_for_path_tk(mode, file_filter, timeout)

plexora/server/utils/pre_normalization.py

@@ -0,0 +1,47 @@
+import numpy as np
+
+
+def preNormalize(input_csvPath, output_csvPath, skip_columns=[]):
+    RAW_DATA = np.genfromtxt(input_csvPath, names=True, dtype=float, delimiter=',')
+    marker_list = RAW_DATA.dtype.names
+    norm_data = RAW_DATA.view((np.float64, len(marker_list)))
+
+    # A list of markers to skip normalization
+    # markers_notToNorm = ['Field_Row', 'Field_Col', 'CellID', 'X_position','Y_position','Percent_Touching','Number_Neighbors','Neighbor_1','Neighbor_2','Neighbor_3','Neighbor_4','Neighbor_5', 'Eccentricity',	'Solidity',	'Extent',	'EulerNumber',	'Perimeter',	'MajorAxisLength',	'MinorAxisLength',	'Orientation',	'X_position',	'Y_position']
+    # A list of markers to skip log10 transform
+    # markers_notToLog = ['DAPI1', 'A488b1', 'A555b1'] # skip 'DAPI1', 'A488b1', 'A555b1'
+    markers_notToLog = []  # nothing to skip
+
+    for marker_id in range(norm_data.shape[1]):
+        if marker_list[marker_id] in skip_columns:
+            continue
+        if marker_list[marker_id] in markers_notToLog:
+            # Log10 transform
+            norm_data[:, marker_id] = np.log10(norm_data[:, marker_id] + 1)
+            print(marker_list[marker_id], 'with log10 transform')
+        else:
+            print(marker_list[marker_id], 'without log10 transform')
+        # Percentile normalization by mapping [0.1%, 99.9%] into [0, 1]
+        min_tile, max_tile = np.percentile(norm_data[:, marker_id], [0.1, 99.9])
+
+        # norm_data[:, marker_id] =(norm_data[:, marker_id] - min_tile) * (max_tile - min_tile) / (max_tile - min_tile) + min_tile
+        norm_data[:, marker_id] = (norm_data[:, marker_id] - min_tile) / (max_tile - min_tile)
+        norm_data[:, marker_id] = np.minimum(norm_data[:, marker_id], 1)
+        norm_data[:, marker_id] = np.maximum(norm_data[:, marker_id], 0)
+
+    with open(output_csvPath, 'w') as f:
+        for marker_id, marker_name in enumerate(marker_list):
+            f.write(marker_name)
+            if marker_id != (len(marker_list) - 1):
+                f.write(',')
+        f.write('\n')
+        for norm_row in norm_data:
+            for elem_id, norm_elem in enumerate(norm_row):
+                f.write(str(norm_elem))
+                if elem_id != (norm_row.shape[0] - 1):
+                    f.write(',')
+            f.write('\n')
+
+# input_csvPath = 'Sample_23.csv'
+# output_csvPath = 'Sample_23_norm2.csv'
+# preNormalize(input_csvPath, output_csvPath)