Didn't need rule order, just a carefully written branch

workflow/playpen.smk

@@ -8,20 +8,18 @@ rule a2b:
     output: "{path}.b"
     shell: "sed -e 's/a/b/g' {wildcards.path}.a >{wildcards.path}.b"
 
-def wildcard_path_endswith_gz(wildcards):
-    return wildcards["path"].endswith(".gz")
-
 rule gunzip:
     input:
-        branch(wildcard_path_endswith_gz,
-               then=[],
-               otherwise="{path}.gz")
+        branch(lambda wildcards: not wildcards["path"].endswith(".gz"),
+               then="{path}.gz",
+               otherwise="/N/A")
     output: "{path}"
     shell: "gunzip -k {wildcards.path}.gz"
 
 rule gzip:
-    input: "{path}"
+    input:
+        branch(lambda wildcards: not wildcards["path"].endswith(".gz"),
+               then="{path}",
+               otherwise="/N/A")
     output: "{path}.gz"
     shell: "gzip -k {wildcards.path}"
-
-ruleorder: a2b > gzip > gunzip