Provide more-specific error details/hints for function lookup failures.

Up to now we've contented ourselves with a one-size-fits-all error
hint when we fail to find any match to a function or procedure call.
That was mostly okay in the beginning, but it was never great, and
since the introduction of named arguments it's really not adequate.
We at least ought to distinguish "function name doesn't exist" from
"function name exists, but not with those argument names".  And the
rules for named-argument matching are arcane enough that some more
detail seems warranted if we match the argument names but the call
still doesn't work.

This patch creates a framework for dealing with these problems:
FuncnameGetCandidates and related code will now pass back a bitmask of
flags showing how far the match succeeded.  This allows a considerable
amount of granularity in the reports.  The set-bits-in-a-bitmask
approach means that when there are multiple candidate functions, the
report will reflect the match(es) that got the furthest, which seems
correct.  Also, we can avoid mentioning "maybe add casts" unless
failure to match argument types is actually the issue.

Extend the same return-a-bitmask approach to OpernameGetCandidates.
The issues around argument names don't apply to operator syntax,
but it still seems worth distinguishing between "there is no
operator of that name" and "we couldn't match the argument types".

While at it, adjust these messages and related ones to more strictly
separate "detail" from "hint", following our message style guidelines'
distinction between those.

Reported-by: Dominique Devienne <ddevienne@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://postgr.es/m/1756041.1754616558@sss.pgh.pa.us

2 files changed, 9 insertions, 8 deletions

diff --git a/doc/src/sgml/sources.sgml b/doc/src/sgml/sources.sgml
index 261f19b3534..760f9b69d47 100644
--- a/doc/src/sgml/sources.sgml
+++ b/doc/src/sgml/sources.sgml
@@ -153,11 +153,12 @@ ereport(ERROR,
         errmsg("function %s is not unique",
                func_signature_string(funcname, nargs,
                                      NIL, actual_arg_types)),
-        errhint("Unable to choose a best candidate function. "
-                "You might need to add explicit typecasts."));
+        errdetail("Could not choose a best candidate function."),
+        errhint("You might need to add explicit type casts."));
 </programlisting>
     This illustrates the use of format codes to embed run-time values into
-    a message text.  Also, an optional <quote>hint</quote> message is provided.
+    a message text.  Also, optional <quote>detail</quote>
+    and <quote>hint</quote> messages are provided.
     The auxiliary function calls can be written in any order, but
     conventionally <function>errcode</function>
     and <function>errmsg</function> appear first.
diff --git a/doc/src/sgml/typeconv.sgml b/doc/src/sgml/typeconv.sgml
index 28748742486..1707bd884dc 100644
--- a/doc/src/sgml/typeconv.sgml
+++ b/doc/src/sgml/typeconv.sgml
@@ -465,9 +465,9 @@ try a similar case with <literal>~</literal>, we get:
 <screen>
 SELECT ~ '20' AS "negation";
 
-ERROR:  operator is not unique: ~ "unknown"
-HINT:  Could not choose a best candidate operator. You might need to add
-explicit type casts.
+ERROR:  operator is not unique: ~ unknown
+DETAIL:  Could not choose a best candidate operator.
+HINT:  You might need to add explicit type casts.
 </screen>
 This happens because the system cannot decide which of the several
 possible <literal>~</literal> operators should be preferred.  We can help
@@ -901,8 +901,8 @@ the parser will try to convert that to <type>text</type>:
 <screen>
 SELECT substr(1234, 3);
 ERROR:  function substr(integer, integer) does not exist
-HINT:  No function matches the given name and argument types. You might need
-to add explicit type casts.
+DETAIL:  No function of that name accepts the given argument types.
+HINT:  You might need to add explicit type casts.
 </screen>
 
 This does not work because <type>integer</type> does not have an implicit cast