Turn on caching everywhere, add logging

A variety of caching issues found by running
all tests with statement caching turned on.

The cache system now has a more conservative approach where
any subclass of a SQL element will by default invalidate
the cache key unless it adds the flag inherit_cache=True
at the class level, or if it implements its own caching.

Add working caching to a few elements that were
omitted previously; fix some caching implementations
to suit lesser used edge cases such as json casts
and array slices.

Refine the way BaseCursorResult and CursorMetaData
interact with caching; to suit cases like Alembic
modifying table structures, don't cache the
cursor metadata if it were created against a
cursor.description using non-positional matching,
e.g. "select *".   if a table re-ordered its columns
or added/removed, now that data is obsolete.

Additionally we have to adapt the cursor metadata
_keymap regardless of if we just processed
cursor.description, because if we ran against
a cached SQLCompiler we won't have the right
columns in _keymap.

Other refinements to how and when we do this
adaption as some weird cases
were exposed in the Postgresql dialect,
a text() construct that names just one column that
is not actually in the statement.   Fixed that
also as it looks like a cut-and-paste artifact
that doesn't actually affect anything.

Various issues with re-use of compiled result maps
and cursor metadata in conjunction with tables being
changed, such as change in order of columns.

mappers can be cleared but the class remains, meaning
a mapper has to use itself as the cache key not the class.

lots of bound parameter / literal issues, due to Alembic
creating a straight subclass of bindparam that renders
inline directly.   While we can update Alembic to not
do this, we have to assume other people might be doing
this, so bindparam() implements the inherit_cache=True
logic as well that was a bit involved.

turn on cache stats in logging.

Includes a fix to subqueryloader which moves all setup to
the create_row_processor() phase and elminates any storage
within the compiled context.   This includes some changes
to create_row_processor() signature and a revising of the
technique used to determine if the loader can participate
in polymorphic queries, which is also applied to
selectinloading.

DML update.values() and ordered_values() now coerces the
keys as we have tests that pass an arbitrary class here
which only includes __clause_element__(), so the
key can't be cached unless it is coerced.  this in turn
changed how composite attributes support bulk update
to use the standard approach of ClauseElement with
annotations that are parsed in the ORM context.

memory profiling successfully caught that the Session
from Query was getting passed into _statement_20()
so that was a big win for that test suite.

Apparently Compiler had .execute() and .scalar() methods
stuck on it, these date back to version 0.4 and there
was a single test in the PostgreSQL dialect tests
that exercised it for no apparent reason.   Removed
these methods as well as the concept of a Compiler
holding onto a "bind".

Fixes: #5386

Change-Id: I990b43aab96b42665af1b2187ad6020bee778784

1 files changed, 54 insertions, 35 deletions

diff --git a/lib/sqlalchemy/engine/cursor.py b/lib/sqlalchemy/engine/cursor.py
index d03d79df7..abffe0d1f 100644
--- a/lib/sqlalchemy/engine/cursor.py
+++ b/lib/sqlalchemy/engine/cursor.py
@@ -51,6 +51,7 @@ class CursorResultMetaData(ResultMetaData):
         "_keys",
         "_tuplefilter",
         "_translated_indexes",
+        "_safe_for_cache"
         # don't need _unique_filters support here for now.  Can be added
         # if a need arises.
     )
@@ -104,11 +105,11 @@ class CursorResultMetaData(ResultMetaData):
         return new_metadata
 
     def _adapt_to_context(self, context):
-        """When using a cached result metadata against a new context,
-        we need to rewrite the _keymap so that it has the specific
-        Column objects in the new context inside of it.  this accommodates
-        for select() constructs that contain anonymized columns and
-        are cached.
+        """When using a cached Compiled construct that has a _result_map,
+        for a new statement that used the cached Compiled, we need to ensure
+        the keymap has the Column objects from our new statement as keys.
+        So here we rewrite keymap with new entries for the new columns
+        as matched to those of the cached statement.
 
         """
         if not context.compiled._result_columns:
@@ -124,14 +125,15 @@ class CursorResultMetaData(ResultMetaData):
         # to the result map.
         md = self.__class__.__new__(self.__class__)
 
-        md._keymap = self._keymap.copy()
+        md._keymap = dict(self._keymap)
 
         # match up new columns positionally to the result columns
         for existing, new in zip(
             context.compiled._result_columns,
             invoked_statement._exported_columns_iterator(),
         ):
-            md._keymap[new] = md._keymap[existing[RM_NAME]]
+            if existing[RM_NAME] in md._keymap:
+                md._keymap[new] = md._keymap[existing[RM_NAME]]
 
         md.case_sensitive = self.case_sensitive
         md._processors = self._processors
@@ -147,6 +149,7 @@ class CursorResultMetaData(ResultMetaData):
         self._tuplefilter = None
         self._translated_indexes = None
         self.case_sensitive = dialect.case_sensitive
+        self._safe_for_cache = False
 
         if context.result_column_struct:
             (
@@ -341,6 +344,10 @@ class CursorResultMetaData(ResultMetaData):
             self._keys = [elem[0] for elem in result_columns]
             # pure positional 1-1 case; doesn't need to read
             # the names from cursor.description
+
+            # this metadata is safe to cache because we are guaranteed
+            # to have the columns in the same order for new executions
+            self._safe_for_cache = True
             return [
                 (
                     idx,
@@ -359,9 +366,12 @@ class CursorResultMetaData(ResultMetaData):
                 for idx, rmap_entry in enumerate(result_columns)
             ]
         else:
+
             # name-based or text-positional cases, where we need
             # to read cursor.description names
+
             if textual_ordered:
+                self._safe_for_cache = True
                 # textual positional case
                 raw_iterator = self._merge_textual_cols_by_position(
                     context, cursor_description, result_columns
@@ -369,6 +379,9 @@ class CursorResultMetaData(ResultMetaData):
             elif num_ctx_cols:
                 # compiled SQL with a mismatch of description cols
                 # vs. compiled cols, or textual w/ unordered columns
+                # the order of columns can change if the query is
+                # against a "select *", so not safe to cache
+                self._safe_for_cache = False
                 raw_iterator = self._merge_cols_by_name(
                     context,
                     cursor_description,
@@ -376,7 +389,9 @@ class CursorResultMetaData(ResultMetaData):
                     loose_column_name_matching,
                 )
             else:
-                # no compiled SQL, just a raw string
+                # no compiled SQL, just a raw string, order of columns
+                # can change for "select *"
+                self._safe_for_cache = False
                 raw_iterator = self._merge_cols_by_none(
                     context, cursor_description
                 )
@@ -1152,7 +1167,6 @@ class BaseCursorResult(object):
 
     out_parameters = None
     _metadata = None
-    _metadata_from_cache = False
     _soft_closed = False
     closed = False
 
@@ -1209,33 +1223,38 @@ class BaseCursorResult(object):
     def _init_metadata(self, context, cursor_description):
         if context.compiled:
             if context.compiled._cached_metadata:
-                cached_md = self.context.compiled._cached_metadata
-                self._metadata_from_cache = True
-
-                # result rewrite/ adapt step.  two translations can occur here.
-                # one is if we are invoked against a cached statement, we want
-                # to rewrite the ResultMetaData to reflect the column objects
-                # that are in our current selectable, not the cached one.  the
-                # other is, the CompileState can return an alternative Result
-                # object.   Finally, CompileState might want to tell us to not
-                # actually do the ResultMetaData adapt step if it in fact has
-                # changed the selected columns in any case.
-                compiled = context.compiled
-                if (
-                    compiled
-                    and not compiled._rewrites_selected_columns
-                    and compiled.statement is not context.invoked_statement
-                ):
-                    cached_md = cached_md._adapt_to_context(context)
-
-                self._metadata = metadata = cached_md
-
+                metadata = self.context.compiled._cached_metadata
             else:
-                self._metadata = (
-                    metadata
-                ) = context.compiled._cached_metadata = self._cursor_metadata(
-                    self, cursor_description
-                )
+                metadata = self._cursor_metadata(self, cursor_description)
+                if metadata._safe_for_cache:
+                    context.compiled._cached_metadata = metadata
+
+            # result rewrite/ adapt step.  this is to suit the case
+            # when we are invoked against a cached Compiled object, we want
+            # to rewrite the ResultMetaData to reflect the Column objects
+            # that are in our current SQL statement object, not the one
+            # that is associated with the cached Compiled object.
+            # the Compiled object may also tell us to not
+            # actually do this step; this is to support the ORM where
+            # it is to produce a new Result object in any case, and will
+            # be using the cached Column objects against this database result
+            # so we don't want to rewrite them.
+            #
+            # Basically this step suits the use case where the end user
+            # is using Core SQL expressions and is accessing columns in the
+            # result row using row._mapping[table.c.column].
+            compiled = context.compiled
+            if (
+                compiled
+                and compiled._result_columns
+                and context.cache_hit
+                and not compiled._rewrites_selected_columns
+                and compiled.statement is not context.invoked_statement
+            ):
+                metadata = metadata._adapt_to_context(context)
+
+            self._metadata = metadata
+
         else:
             self._metadata = metadata = self._cursor_metadata(
                 self, cursor_description