From c0d3f379053d3d95cbac5ca6539101e8b68d59b3 Mon Sep 17 00:00:00 2001
From: Tony Garnock-Jones <tonyg@leastfixedpoint.com>
Date: Fri, 3 Jun 2022 21:13:59 +0200
Subject: [PATCH] python: merge(), and support for and-pattern unparsing

---
 implementations/python/preserves/__init__.py |  2 +
 implementations/python/preserves/merge.py    | 42 ++++++++++++++++++++
 implementations/python/preserves/schema.py   |  4 +-
 3 files changed, 47 insertions(+), 1 deletion(-)
 create mode 100644 implementations/python/preserves/merge.py

diff --git a/implementations/python/preserves/__init__.py b/implementations/python/preserves/__init__.py
index 7a163ef..3ec1b12 100644
--- a/implementations/python/preserves/__init__.py
+++ b/implementations/python/preserves/__init__.py
@@ -6,6 +6,8 @@ from .error import DecodeError, EncodeError, ShortPacket
 from .binary import Decoder, Encoder, decode, decode_with_annotations, encode, canonicalize
 from .text import Parser, Formatter, parse, parse_with_annotations, stringify
 
+from .merge import merge
+
 from . import fold
 
 loads = parse
diff --git a/implementations/python/preserves/merge.py b/implementations/python/preserves/merge.py
new file mode 100644
index 0000000..c04fac1
--- /dev/null
+++ b/implementations/python/preserves/merge.py
@@ -0,0 +1,42 @@
+from .values import ImmutableDict, dict_kvs, Embedded, Record
+
+def merge_embedded_id(a, b):
+    return a if a is b else None
+
+def merge(v0, *vs, merge_embedded=None):
+    v = v0
+    for vN in vs:
+        v = merge2(v, vN, merge_embedded=merge_embedded)
+    return v
+
+def _die():
+    raise ValueError('Cannot merge items')
+
+def merge_seq(aa, bb, merge_embedded=None):
+    if len(aa) != len(bb): _die()
+    return [merge2(a, b, merge_embedded=merge_embedded) for (a, b) in zip(aa, bb)]
+
+def merge2(a, b, merge_embedded=None):
+    if a == b:
+        return a
+    if isinstance(a, (list, tuple)) and isinstance(b, (list, tuple)):
+        return merge_seq(a, b)
+    if isinstance(a, (set, frozenset)) and isinstance(b, (set, frozenset)):
+        _die()
+    if isinstance(a, dict) and isinstance(b, dict):
+        r = {}
+        for (ak, av) in a.items():
+            bv = b.get(ak, None)
+            r[ak] = av if bv is None else merge2(av, bv, merge_embedded=merge_embedded)
+        for (bk, bv) in b.items():
+            if bk not in r:
+                r[bk] = bv
+        return r
+    if isinstance(a, Record) and isinstance(b, Record):
+        return Record(merge2(a.key, b.key, merge_embedded=merge_embedded),
+                      merge_seq(a.fields, b.fields, merge_embedded=merge_embedded))
+    if isinstance(a, Embedded) and isinstance(b, Embedded):
+        m = (merge_embedded or merge_embedded_id)(a.embeddedValue, b.embeddedValue)
+        if m is None: _die()
+        return Embedded(m)
+    _die()
diff --git a/implementations/python/preserves/schema.py b/implementations/python/preserves/schema.py
index becdb23..a94f82c 100644
--- a/implementations/python/preserves/schema.py
+++ b/implementations/python/preserves/schema.py
@@ -319,7 +319,9 @@ def encode(p, v):
         return tuple(encode(pp, v) for pp in p[0]) + encode(p[1], v)
     if p.key == DICT:
         return dict((k, encode(pp, v)) for (k, pp) in p[0].items())
-    raise ValueError('Bad schema')
+    if p.key == AND:
+        return merge(*[encode(pp, v) for pp in p[0]])
+    raise ValueError(f'Bad schema {p}')
 
 def module_path_str(mp):
     return '.'.join([e.name for e in mp])