From b07e7bfa2067b01642a5627dfaf15e3ebd77dec6 Mon Sep 17 00:00:00 2001
From: tristanpwdennis <tristanpwdennis@gmail.com>
Date: Wed, 18 Jun 2025 17:53:17 -0500
Subject: [PATCH 1/2] Cherry-pick: add hashing to anopheles.py from 1b196344

---
 malariagen_data/anopheles.py | 71 +++++++++++++++++++++++++++---------
 1 file changed, 54 insertions(+), 17 deletions(-)

diff --git a/malariagen_data/anopheles.py b/malariagen_data/anopheles.py
index ed9cb65cf..7fc78451e 100644
--- a/malariagen_data/anopheles.py
+++ b/malariagen_data/anopheles.py
@@ -574,31 +574,68 @@ def roh_hmm(
         debug = self._log.debug
 
         resolved_region: Region = parse_single_region(self, region)
-        del region
 
-        debug("compute windowed heterozygosity")
-        sample_id, sample_set, windows, counts = self._sample_count_het(
+        name = "roh"
+
+        params = dict(
             sample=sample,
-            region=resolved_region,
-            site_mask=site_mask,
+            region=region,
             window_size=window_size,
+            site_mask=site_mask,
             sample_set=sample_set,
-            chunks=chunks,
-            inline_array=inline_array,
-        )
-
-        debug("compute runs of homozygosity")
-        df_roh = self._roh_hmm_predict(
-            windows=windows,
-            counts=counts,
             phet_roh=phet_roh,
             phet_nonroh=phet_nonroh,
             transition=transition,
-            window_size=window_size,
-            sample_id=sample_id,
-            contig=resolved_region.contig,
+            chunks=chunks,
+            inline_array=inline_array,
         )
 
+        del region
+
+        try:
+            # Load cached numeric data, adding str / obj data again.
+            results = self.results_cache_get(name=name, params=params)
+            df_roh = pd.DataFrame(results)
+            df_roh["sample_id"] = sample
+            df_roh["contig"] = resolved_region.contig
+
+        except CacheMiss:
+            debug("compute windowed heterozygosity")
+            sample_id, sample_set, windows, counts = self._sample_count_het(
+                sample=sample,
+                region=resolved_region,
+                site_mask=site_mask,
+                window_size=window_size,
+                sample_set=sample_set,
+                chunks=chunks,
+                inline_array=inline_array,
+            )
+
+            debug("compute runs of homozygosity")
+            df_roh = self._roh_hmm_predict(
+                windows=windows,
+                counts=counts,
+                phet_roh=phet_roh,
+                phet_nonroh=phet_nonroh,
+                transition=transition,
+                window_size=window_size,
+                sample_id=sample_id,
+                contig=resolved_region.contig,
+            )
+
+            # Specify numeric columns to save (saving obj - sample ID and contig - breaks the save.
+            columns_to_save = [
+                "roh_start",
+                "roh_stop",
+                "roh_length",
+                "roh_is_marginal",
+            ]
+            self.results_cache_set(
+                name=name,
+                params=params,
+                results={col: df_roh[col].to_numpy() for col in columns_to_save},
+            )
+
         return df_roh
 
     @check_types
@@ -1306,7 +1343,7 @@ def ihs_gwss(
     ) -> Tuple[np.ndarray, np.ndarray]:
         # change this name if you ever change the behaviour of this function, to
         # invalidate any previously cached data
-        name = self._ihs_gwss_cache_name
+        name = "roh"
 
         params = dict(
             contig=contig,

From 9ccf96d92459994abfd8b47a85d062d4e97b2dd0 Mon Sep 17 00:00:00 2001
From: tristanpwdennis <tristanpwdennis@gmail.com>
Date: Wed, 18 Jun 2025 18:17:27 -0500
Subject: [PATCH 2/2] Add more comments, tidy up a little

---
 malariagen_data/anopheles.py | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/malariagen_data/anopheles.py b/malariagen_data/anopheles.py
index 7fc78451e..c5411a6b1 100644
--- a/malariagen_data/anopheles.py
+++ b/malariagen_data/anopheles.py
@@ -575,7 +575,8 @@ def roh_hmm(
 
         resolved_region: Region = parse_single_region(self, region)
 
-        name = "roh"
+        # Create params for hashing.
+        name = "roh_v1"
 
         params = dict(
             sample=sample,
@@ -592,8 +593,9 @@ def roh_hmm(
 
         del region
 
+        # The caching struggles with saving variable length strings, so we can just load/save the numeric data, and
+        # add the strings (sample ID and contig) from user input.
         try:
-            # Load cached numeric data, adding str / obj data again.
             results = self.results_cache_get(name=name, params=params)
             df_roh = pd.DataFrame(results)
             df_roh["sample_id"] = sample
@@ -623,13 +625,15 @@ def roh_hmm(
                 contig=resolved_region.contig,
             )
 
-            # Specify numeric columns to save (saving obj - sample ID and contig - breaks the save.
+            # Specify numeric columns to save to cache. (See above - variable length strings can break the save).
             columns_to_save = [
                 "roh_start",
                 "roh_stop",
                 "roh_length",
                 "roh_is_marginal",
             ]
+
+            # Save cache
             self.results_cache_set(
                 name=name,
                 params=params,