Modularize the flu case study

demos/02_flu/.gitignore

@@ -1,5 +1,10 @@
 log
 *pdf
 *rda
-*.json
+tree.json
 __pycache__/
+nexus.py
+pool-cpp.out
+pool.R
+pool.cpp
+pool.py

demos/02_flu/Makefile

@@ -1,21 +1,9 @@
-all:
-	morloc typecheck -r main.loc
-	morloc make main.loc
-	./nexus.py -h
-
-.PHONY:
-search:
-	./nexus.py searchEntrez config.json > searchResult.json
+all: nexus.py
+	./nexus.py makeTree data/config.json > tree.json
+	./nexus.py plot data/config.json tree.json
 
-.PHONY:
-run:
-	./nexus.py classifyByDate config.json > tree.json
-	./nexus.py plotCladeTree config.json tree.json
-
-.PHONY:
-makec:
-	g++ --std=c++17 -o pool-cpp.out pool.cpp -I/home/z/.morloc/include -I/home/z/.morloc/src/morloc/plane/morloclib/bio/tree -I/home/z/.morloc/src/morloc/plane/morloclib/cppbase -I/home/z/.morloc/src/morloc/plane/morloclib/bio/algo -I/home/z/.morloc/src/morloc/plane/morloclib/matrix/eigen
+nexus.py:
+	morloc make main.loc
 
-.PHONY: clean
 clean:
-	rm -rf pool* nexus* *pdf *rda __pycache__
+	rm -rf pool* nexus* *pdf *rda __pycache__ tree.json

demos/02_flu/config.json → demos/02_flu/data/config.json

@@ -1,7 +1,7 @@
 { "mindate" : "2021/01/01"
 , "maxdate" : "2021/01/14"
-, "reffile" : "refs.txt"
+, "reffile" : "data/refs.txt"
 , "treefile" : "tree.pdf"
-, "email" : "zbwrnz@gmail.com"
+, "email" : "wena@mailinator.com"
 , "query" : "Influenza+A+Virus[Organism]+H3N2[ALL]+HA[ALL]"
 }

demos/02_flu/lib/classify/main.loc

@@ -0,0 +1,41 @@
+module lib.classify (classify)
+
+import lib.flutypes (Clade)
+
+-- Import generic functions
+import base
+  ( id
+  , snd
+  , ifelse
+  , eq
+  , size
+  , fst
+  , branch
+  , head
+  , const
+  , unique
+  , ne
+  , filter
+  )
+
+-- Import the tree type and generic functions for operating on it
+import bio.tree
+  ( RootedTree
+  , pullNode
+  , push
+  )
+
+-- Traverse the phylogenetic tree setting clade labels
+classify n e a :: RootedTree n e (a, Clade) -> RootedTree Str e (a, Clade)
+classify
+  = push id passClade setLeaf
+  . pullNode snd pullClade
+  where
+    passClade parent edge child = (edge, ifelse (eq 0 (size child)) parent child)
+    setLeaf parent edge leaf = (edge, (fst leaf, parent))
+    pullClade xs
+      = branch (eq 1 . size) head (const "") seenClades
+      where
+        seenClades = ( unique
+                     . filter (ne 0 . size)
+                     ) xs

demos/02_flu/lib/flutypes/main.loc

@@ -0,0 +1,9 @@
+module lib.flutypes (*)
+
+import types(Str)
+
+-- Type aliases
+type Accession = Str
+type Clade = Str
+type Sequence = Str
+type Date = Str

demos/02_flu/lib/retrieve/entrez.py

@@ -0,0 +1,21 @@
+#  parseRecord :: JsonObj -> (JsonObj, Sequence)
+def parseRecord(jsonObj):
+    sequence = jsonObj["GBSeq_sequence"].upper() 
+    del jsonObj["GBSeq_sequence"]
+    return (jsonObj, sequence)
+
+#  labelRef :: Map Accession Clade -> JsonObj -> (JsonObj, Clade)
+def labelRef(cladeMap, jsonObj):
+    accession = jsonObj["GBSeq_primary-accession"]
+    if accession in cladeMap:
+        return (jsonObj, cladeMap[accession])
+    else:
+        return (jsonObj, "")
+
+# setLeafName :: (JsonObj, Clade) -> Str
+def setLeafName(meta):
+    (jsonObj, clade) = meta
+    return ( clade + "|" +
+             jsonObj["GBSeq_primary-accession"] + "|" +
+             jsonObj["GBSeq_length"]
+           )

demos/02_flu/lib/retrieve/main.loc

@@ -0,0 +1,75 @@
+module lib.retrieve (retrieve, setLeafName, FluConfig)
+
+import lib.flutypes
+ ( Accession
+ , Clade
+ , Date
+ , Sequence
+ )
+
+-- Import functions for accessing public sequence records
+import bio.db
+  ( searchEntrez as fetchIds
+  , nucleotideAccessionToJson as fetchRecords
+  , EntrezSearchConfig
+  , EntrezFetchConfig
+  )
+
+-- Import generic functions
+import base (map, concat, shard, join, keys, onFst, sleep, readMap)
+import json (JsonObj)
+import types (Filename)
+
+
+-- Define the configuration record
+record FluConfig = FluConfig
+  { mindate :: Date
+  , maxdate :: Date
+  , reffile :: Filename
+  , treefile :: Filename
+  , query :: Str
+  , email :: Str
+  }
+
+-- Specify the representation of this record in Python and R
+record Py => FluConfig = "dict"
+record R => FluConfig = "list"
+
+-- Source python functions for dealing with Entrez records
+source Py from "entrez.py"
+  ( "parseRecord"
+  , "labelRef"
+  , "setLeafName"
+  )
+
+-- Source C++ alternative for one of these functions
+source Cpp from "entrez.hpp" ("setLeafName")
+
+-- Define the general type of each function
+parseRecord :: JsonObj -> (JsonObj, Sequence)
+labelRef :: Map Accession Clade -> JsonObj -> (JsonObj, Clade)
+setLeafName :: (JsonObj, Sequence) -> Str
+
+-- Retrieve sequence data from Entrez and tag reference strains
+retrieve :: FluConfig -> [((JsonObj, Clade), Sequence)]
+retrieve config =
+  ( map (onFst (labelRef refmap))
+  . concat
+  . map ( map parseRecord
+        . sleep 1.0
+        . fetchRecords fetchConfig
+        )
+  . shard 30
+  . join (keys refmap)
+  . fetchIds searchConfig
+  ) config@query
+  where
+    searchConfig =
+      { email = config@email
+      , db = "nuccore"
+      , mindate = config@mindate
+      , maxdate = config@maxdate
+      , retmax = 1000
+      }
+    fetchConfig = { email = config@email }
+    refmap = readMap config@reffile

demos/02_flu/lib/treeplot/main.loc

@@ -0,0 +1,7 @@
+module lib.treeplot (plotTree)
+
+import bio.tree(RootedTree)
+import types (Filename)
+
+source R from "plot-tree.R" ("plotTree")
+plotTree :: Filename -> RootedTree Str Real Str -> ()