Fix #232: tempdir is not respected in tl.macs3

docs/changelog.md

@@ -6,6 +6,7 @@
 
 - Fix: #221: 'pp.knn' with 'method=pynndescent' invalid csr matrix.
 - Fix: #226: Backed AnnData does not support numpy string array. 
+- Fix: #232: `tempdir` is not respected in `tl.macs3`.
 
 ## Release 2.5.3 (released Jan 16, 2024)
 

snapatac2-python/snapatac2/tools/_call_peaks.py

@@ -85,23 +85,14 @@ def macs3(
     """
     from MACS3.Signal.PeakDetect import PeakDetect
     from math import log
-    from multiprocess import get_context
-    from tqdm import tqdm
     import tempfile
 
     if isinstance(groupby, str):
         groupby = list(adata.obs[groupby])
     if replicate is not None and isinstance(replicate, str):
         replicate = list(adata.obs[replicate])
-    if tempdir is None:
-        tempdir = Path(tempfile.mkdtemp())
-    else:
-        tempdir = Path(tempfile.mkdtemp(dir=tempdir))
-
-    logging.info("Exporting fragments...")
-    fragments = _snapatac2.export_tags(adata, tempdir, groupby, replicate, max_frag_size, selections)
 
-    # General options
+    # MACS3 options
     options = type('MACS3_OPT', (), {})()
     options.info = lambda _: None
     options.debug = lambda _: None
@@ -130,37 +121,43 @@ def macs3(
     options.d = extsize
     options.scanwindow = 2 * options.d
 
-    def _call_peaks(tags):
-        merged, reps = _snapatac2.create_fwtrack_obj(tags)
-        options.log_qvalue = log(qvalue, 10) * -1
-        logging.getLogger().setLevel(logging.CRITICAL + 1)
-        peakdetect = PeakDetect(treat=merged, opt=options)
-        peakdetect.call_peaks()
-        peakdetect.peaks.filter_fc(fc_low = options.fecutoff)
-        merged = peakdetect.peaks
-
-        others = []
-        if replicate_qvalue is not None:
-            options.log_qvalue = log(replicate_qvalue, 10) * -1
-        for x in reps:
-            peakdetect = PeakDetect(treat=x, opt=options)
+    with tempfile.TemporaryDirectory(dir=tempdir) as tmpdirname:
+        logging.info("Exporting fragments...")
+        fragments = _snapatac2.export_tags(adata, tmpdirname, groupby, replicate, max_frag_size, selections)
+
+        def _call_peaks(tags):
+            import tempfile
+            tempfile.tempdir = tmpdirname  # Overwrite the default tempdir in MACS3
+            merged, reps = _snapatac2.create_fwtrack_obj(tags)
+            options.log_qvalue = log(qvalue, 10) * -1
+            logging.getLogger().setLevel(logging.CRITICAL + 1)
+            peakdetect = PeakDetect(treat=merged, opt=options)
             peakdetect.call_peaks()
             peakdetect.peaks.filter_fc(fc_low = options.fecutoff)
-            others.append(peakdetect.peaks)
-
-        logging.getLogger().setLevel(logging.INFO)
-        return _snapatac2.find_reproducible_peaks(merged, others, blacklist)
-
-    logging.info("Calling peaks...")
-    peaks = _par_map(_call_peaks, [(x,) for x in fragments.values()], n_jobs)
-    peaks = {k: v for k, v in zip(fragments.keys(), peaks)}
-    if inplace:
-        if adata.isbacked:
-            adata.uns[key_added] = peaks
+            merged = peakdetect.peaks
+
+            others = []
+            if replicate_qvalue is not None:
+                options.log_qvalue = log(replicate_qvalue, 10) * -1
+            for x in reps:
+                peakdetect = PeakDetect(treat=x, opt=options)
+                peakdetect.call_peaks()
+                peakdetect.peaks.filter_fc(fc_low = options.fecutoff)
+                others.append(peakdetect.peaks)
+
+            logging.getLogger().setLevel(logging.INFO)
+            return _snapatac2.find_reproducible_peaks(merged, others, blacklist)
+
+        logging.info("Calling peaks...")
+        peaks = _par_map(_call_peaks, [(x,) for x in fragments.values()], n_jobs)
+        peaks = {k: v for k, v in zip(fragments.keys(), peaks)}
+        if inplace:
+            if adata.isbacked:
+                adata.uns[key_added] = peaks
+            else:
+                adata.uns[key_added] = {k: v.to_pandas() for k, v in peaks.items()}
         else:
-            adata.uns[key_added] = {k: v.to_pandas() for k, v in peaks.items()}
-    else:
-        return peaks
+            return peaks
 
 def merge_peaks(
     peaks: dict[str, 'polars.DataFrame'],