Re: [PATCH] Add pg_get_database_ddl() function to reconstruct CREATE DATABASE statement

On Mon, 9 Mar 2026 at 17:08, Akshay Joshi <akshay(dot)joshi(at)enterprisedb(dot)com> wrote:
>
> I have resolved the review comments. Removed the hardcoded 'utf-8' encoding, the logic now retrieves the encoding dynamically. This ensures accuracy because the default encoding for CREATE DATABASE is derived from the template database (template1) rather than being a static value.
>
> Attached is the v13 patch, now ready for further review.
>
> On Fri, Mar 6, 2026 at 8:13 PM Japin Li <japinli(at)hotmail(dot)com> wrote:
>>
>>
>> Hi, Akshay
>>
>> On Fri, 06 Mar 2026 at 17:21, Akshay Joshi <akshay(dot)joshi(at)enterprisedb(dot)com> wrote:
>> > I have updated the documentation.
>> > Attached is the v12 patch, now ready for further review.
>> >
>>
>> Thanks for updating the patch.
>>
>> I might not have expressed myself clearly in my last email — apologies.
>> Here's a diff that should make it clearer.
>>
>> diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
>> index bbf82c1d6c0..1ed56ee71ab 100644
>> --- a/src/backend/utils/adt/ruleutils.c
>> +++ b/src/backend/utils/adt/ruleutils.c
>> @@ -13859,7 +13859,6 @@ parse_ddl_options(FunctionCallInfo fcinfo, int variadic_start)
>> bool *nulls;
>> Oid *types;
>> int nargs;
>> - bool found = false;
>>
>> /* Extract variadic arguments */
>> nargs = extract_variadic_args(fcinfo, variadic_start, true,
>> @@ -13887,8 +13886,7 @@ parse_ddl_options(FunctionCallInfo fcinfo, int variadic_start)
>> {
>> char *name;
>> bool bval;
>> -
>> - found = false;
>> + bool found = false;
>>
>> /* Key must not be null */
>> if (nulls[i])
>> @@ -13925,8 +13923,8 @@ parse_ddl_options(FunctionCallInfo fcinfo, int variadic_start)
>> if (!parse_bool(valstr, &bval))
>> ereport(ERROR,
>> (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
>> - errmsg("argument %d: invalid value \"%s\" for key \"%s\"",
>> - i + 2, valstr, name),
>> + errmsg("value for option \"%s\" at position %d has invalid value \"%s\"",
>> + name, i + 2, valstr),
>> errhint("Valid values are: true, false, yes, no, 1, 0, on, off.")));
>> pfree(valstr);
>> }
>> @@ -13934,8 +13932,8 @@ parse_ddl_options(FunctionCallInfo fcinfo, int variadic_start)
>> {
>> ereport(ERROR,
>> (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
>> - errmsg("argument %d: value for key \"%s\" must be boolean or text type",
>> - i + 2, name)));
>> + errmsg("value for option \"%s\" at position %d has type %s, expected type boolean or text",
>> + name, i + 2, format_type_be(types[i + 1]))));
>> }
>>
>> /*
>> @@ -13983,7 +13981,7 @@ parse_ddl_options(FunctionCallInfo fcinfo, int variadic_start)
>> /*
>> * pg_get_database_ddl
>> *
>> - * Generate a CREATE DATABASE statement for the specified database name or oid.
>> + * Generate a CREATE DATABASE statement for the specified database oid.
>> *
>> * db_oid - OID of the database for which to generate the DDL.
>> * options - Variadic name/value pairs to modify the output.
>>
>> --
>> Regards,
>> Japin Li
>> ChengDu WenWu Information Technology Co., Ltd.

Hi!
I noticed this in v13: DDLOptionDef is missing from typedefs.list.
This results in pgident to incorrectly format sources.

--
Best regards,
Kirill Reshke