在PLPGSQL函数中，我有一些行来检查记录是否存在。好吧，即使没有记录，FOUND变量也会设置为TRUE。非常奇怪。
我正在测试的行是：
query:='SELECT i.identity_id FROM app.identity as i,' || quote_ident(schema_name) || '.users as u,' || quote_ident(schema_name) || '.udata as ud WHERE i.identity_id=' || p_identity_id || ' and u.identity_id=i.identity_id and ud.identity_id=i.identity_id';
RAISE WARNING 'query=%',query;
PERFORM query;
RAISE WARNING 'FOUND=%',FOUND;
IF FOUND
THEN
RAISE WARNING 'Record found, rising an exception';
RAISE EXCEPTION USING ERRCODE = 'AA002';
END IF;
这是输出：
WARNING: query=SELECT i.identity_id FROM app.identity as i,"comp-1049007476".users as u,"comp-1049007476".udata as ud WHERE i.identity_id=-1615382132 and u.identity_id=i.identity_id and ud.identity_id=i.identity_id
WARNING: FOUND=t
WARNING: Record found, rising an exception
这是pg_log：
2017-01-09 09:47:21.906 CST > LOG: execute <unnamed>: SELECT app.create_identity($1::varchar,'') as identity_id
< 2017-01-09 09:47:21.906 CST > DETAIL: parameters: $1 = 'someuser@domain.com'
< 2017-01-09 09:47:21.908 CST > WARNING: new_identity_id=-1615382132
< 2017-01-09 09:47:21.908 CST > CONTEXT: PL/pgSQL function app.create_identity(character varying,character varying) line 18 at RAISE
< 2017-01-09 09:47:21.917 CST > LOG: execute <unnamed>: select app.add_user(
$1::integer,
$2::varchar,
$3::integer,
$4::varchar,
$5::varchar,
$6::varchar,
$7::varchar,
$8::varchar,
$9::varchar,
$10::varchar,
$11::integer,
$12::integer,
$13::integer,
$14::integer,
$15::integer,
$16::integer,
$17::smallint,
$18::boolean,
$19::boolean,
$20::boolean,
$21::boolean,
$22::boolean
)
< 2017-01-09 09:47:21.917 CST > DETAIL: parameters: $1 = '905220468', $2 = '763715373817831', $3 = '-1049007476', $4 = 'Some User', $5 = '44444', $6 = '', $7 = '', $8 = '', $9 = '', $10 = '', $11 = '-1615382132', $12 = '0', $13 = '1', $14 = '0', $15 = '0', $16 = '0', $17 = '0', $18 = 't', $19 = 'f', $20 = 'f', $21 = 'f', $22 = 'f'
< 2017-01-09 09:47:21.919 CST > WARNING: query=SELECT i.identity_id FROM app.identity as i,"comp-1049007476".users as u,"comp-1049007476".udata as ud WHERE i.identity_id=-1615382132 and u.identity_id=i.identity_id and ud.identity_id=i.identity_id
< 2017-01-09 09:47:21.919 CST > CONTEXT: PL/pgSQL function app.add_user(integer,character varying,integer,character varying,character varying,character varying,character varying,character varying,character varying,character varying,integer,integer,integer,integer,integer,integer,smallint,boolean,boolean,boolean,boolean,boolean) line 25 at RAISE
< 2017-01-09 09:47:21.919 CST > WARNING: FOUND=t
< 2017-01-09 09:47:21.919 CST > CONTEXT: PL/pgSQL function app.add_user(integer,character varying,integer,character varying,character varying,character varying,character varying,character varying,character varying,character varying,integer,integer,integer,integer,integer,integer,smallint,boolean,boolean,boolean,boolean,boolean) line 27 at RAISE
< 2017-01-09 09:47:21.919 CST > WARNING: Record found, rising an exception
< 2017-01-09 09:47:21.919 CST > CONTEXT: PL/pgSQL function app.add_user(integer,character varying,integer,character varying,character varying,character varying,character varying,character varying,character varying,character varying,integer,integer,integer,integer,integer,integer,smallint,boolean,boolean,boolean,boolean,boolean) line 30 at RAISE
< 2017-01-09 09:47:21.919 CST > ERROR: Identity already registered
为什么我确定记录不存在？好吧，我把我知道的值放在表格中，但为了以防万一，我确认：
dev=> SELECT i.identity_id FROM app.identity as i,"comp-1049007476".users as u,"comp-1049007476".udata as ud WHERE i.identity_id=-1615382132 and u.identity_id=i.identity_id and ud.identity_id=i.identity_id;
identity_id
-------------
(0 rows)
dev=>
参见，0记录。所以，大问题是，为什么FOUND变量谎言？这里发生了什么？
这是我的功能的完整代码：
CREATE OR REPLACE FUNCTION app.add_user(sess_identity_id int,session_str varchar,sess_company_id int,
p_full_name varchar,
p_mob_phone varchar,
p_home_phone varchar,
p_work_phone varchar,
p_phone_ext varchar,
p_position varchar,
p_notes varchar,
p_identity_id integer,
p_user_id integer,
p_group_id integer,
p_shift_id integer,
p_boss_id integer,
p_crreated_by integer,
p_timezone smallint,
p_utype boolean,
p_create_root boolean,
p_has_subord boolean,
p_inactive boolean,
p_generic boolean
)
RETURNS integer as $$
DECLARE
v_session bigint;
schema_name varchar;
query varchar;
current_ts integer;
v_user_id integer;
v_generic integer;
v_udata_id integer;
_c text;
BEGIN
SELECT extract(epoch from now())::integer into current_ts;
SELECT session_str::bigint INTO v_session;
schema_name:='comp' || sess_company_id;
IF NOT EXISTS (
SELECT 1 FROM app.session WHERE app.session.identity_id=sess_identity_id AND app.session.session=v_session
) THEN
RAISE EXCEPTION USING ERRCODE = 'AA001';
END IF;
query:='SELECT i.identity_id FROM app.identity as i,' || quote_ident(schema_name) || '.users as u,' || quote_ident(schema_name) || '.udata as ud WHERE i.identity_id=' || p_identity_id || ' and u.identity_id=i.identity_id and ud.identity_id=i.identity_id';
RAISE WARNING 'query=%',query;
PERFORM query;
RAISE WARNING 'FOUND=%',FOUND;
IF FOUND
THEN
RAISE WARNING 'Record found, rising an exception';
RAISE EXCEPTION USING ERRCODE = 'AA002';
END IF;
query:='SELECT u.user_id,u.generic FROM ' || quote_ident(schema_name) || '.users as u,identity as i WHERE i.email=p_email AND i.identity_id=u.identity_id';
RAISE WARNING 'query=%',query;
EXECUTE query INTO v_user_id,v_generic;
IF NOT FOUND
THEN
EXECUTE 'INSERT INTO $1.users(
identity_id,
group_id,
shift_id,
boss_id,
created_by,
date_inserted
timezone,
utype,
create_root,
has_subord,
inactive,
generic
) VALUES ($2,$3,$4,$5,$6,$7,$8,$9,$10,$11,$12,$13) RETURNING user_id'
INTO v_user_id
USING schema_name,p _identity_id,p_group_id,p_shift_id,p_boss_id,p_created_by,p_date_inserted,p_timezone,p_utype,p_create_root,p_has_subord,p_inactive,p_generc;
END IF;
if NOT v_generic THEN
RAISE EXCEPTION USING ERRCODE = 'AA003';
END IF;
EXECUTE 'INSERT INTO $1.udata(user_id,identity_id,date_insert,full_name,mob_phone,home_phone,work_phone,phone_ext,position,notes) VALUES ($2,$3,$4,$5,$6,$7,$8,$9,$10,$11)'
INTO v_udata_id
USING schema_name,v_user_id,p_identity_id,current_ts,p_full_name,p_mob_phone,p_home_phone,p_work_phone,p_phone_ext,p_position,p_notes;
RETURN v_udata_id;
EXCEPTION
WHEN SQLSTATE 'AA001' THEN
RAISE EXCEPTION USING ERRCODE = 'AA001', message = 'Invalid session';
WHEN SQLSTATE 'AA002' THEN
RAISE EXCEPTION USING ERRCODE = 'AA002', message = 'Identity already registered';
WHEN SQLSTATE 'AA003' THEN
RAISE EXCEPTION USING ERRCODE = 'AA003', message = 'Not a generic user, can not add email address';
WHEN OTHERS THEN
RAISE NOTICE 'add_user() failed with... error: % %',SQLSTATE,SQLERRM;
GET STACKED DIAGNOSTICS _c = PG_EXCEPTION_CONTEXT;
RAISE NOTICE 'context: >>%<<', _c;
RAISE EXCEPTION USING MESSAGE = 'An error which is not handled by function';
END
$$ LANGUAGE plpgsql;
表结构，以防万一：
dev=> \d+ app.identity;
Table "app.identity"
Column | Type | Modifiers | Storage | Stats target | Description
---------------+------------------------+-----------+----------+--------------+-------------
identity_id | integer | not null | plain | |
created_by | integer | | plain | |
email | character varying(128) | | extended | |
date_inserted | integer | | plain | |
password | character varying(32) | | extended | |
validated | smallint | | plain | |
Indexes:
"identity_pkey" PRIMARY KEY, btree (identity_id)
"identity_email_key" UNIQUE CONSTRAINT, btree (email)
dev=> \d+ "comp-1049007476".users;
Table "comp-1049007476.users"
Column | Type | Modifiers | Storage | Stats target | Description
---------------+----------+---------------------------------------------------------------------------+---------+--------------+-------------
user_id | integer | not null default nextval('"comp-1049007476".users_user_id_seq'::regclass) | plain | |
identity_id | integer | default 0 | plain | |
group_id | integer | default 0 | plain | |
shift_id | integer | default 0 | plain | |
boss_id | integer | default 0 | plain | |
created_by | integer | default 0 | plain | |
date_inserted | integer | default 0 | plain | |
timezone | smallint | default 0 | plain | |
utype | boolean | default false | plain | |
create_root | boolean | default false | plain | |
has_subord | boolean | default false | plain | |
inactive | boolean | default false | plain | |
generic | boolean | default false | plain | |
dev-> \d+ "comp-1049007476".udata
Table "comp-1049007476.udata"
Column | Type | Modifiers | Storage | Stats target | Description
---------------+------------------------+----------------------------------------------------------------------------+----------+--------------+-------------
udata_id | integer | not null default nextval('"comp-1049007476".udata_udata_id_seq'::regclass) | plain | |
user_id | integer | default 0 | plain | |
identity_id | integer | default 0 | plain | |
date_inserted | integer | default 0 | plain | |
full_name | character varying(64) | default ''::character varying | extended | |
mob_phone | character varying(16) | default ''::character varying | extended | |
home_phone | character varying(16) | default ''::character varying | extended | |
work_phone | character varying(16) | default ''::character varying | extended | |
phone_ext | character varying(8) | default ''::character varying | extended | |
position | character varying(64) | default ''::character varying | extended | |
notes | character varying(128) | default ''::character varying | extended | |
dev->
答案 0 :(得分：6)
基本问题在于使用PERFORM语句。此语句专为无需结果处理的功能评估而设计。它不是为动态SQL设计的（其中SQL以字符串形式输入）。
PLpgSQL几乎将所有语句翻译为SELECT。 PERFORM也不例外。
PERFORM fx(10,20);
被翻译为：
SELECT fx(10,20);
您的代码：
variable := 'SELECT * FROM foo';
PERFORM variable;
被翻译为：
SELECT 'SELECT * FROM foo'; -- it is same like SELECT 'hello';
此查询返回一行，变量FOUND应为true。所以没有什么是奇怪的。
在PLpgSQL中，您应该在静态SQL和动态SQL之间存在很大差异。动态SQL（在运行时汇编查询）仅受EXECUTE语句支持。
使用PERFORM variable是对此语句的简单错误用法 - 但它会生成有效的SQL语句，可以在没有运行时错误的情况下进行评估。