PGSQL函數在特定列的分隔符上拆分表格

Question

我目前正在將一些數據庫從excel格式遷移到PostgreSQL。其中一些記錄在一行中，由分隔符分隔。我想寫由選定的隔膜通過以下方式分離所選列轉換表的功能：

create table janek.temp (a integer, b text); 
insert into janek.temp values (1, 'cat'); 
insert into janek.temp values (2, 'dog;hound'); 

此表中包含2行，我想有一個功能，當我執行：

select * from janek.split_table ('janek', 'temp', 'b', ';') 

我得到3行返回：

我與strugge

1;'cat' 
2;'dog' 
2;'hound' 

目前的問題是 - 不知道如何通過模式名和表名到RETURNS setof $1.$2 AS - 首次執行的查詢不起作用，原因我不明白，錯誤粘貼在下面

我希望我的邏輯對你們很清楚。我提供了示例數據和期望的結果。我對PGSQL功能並不熟悉，但我希望我編寫的代碼易於操作。

我希望這個函數能被重用，對我的數據遷移似乎很有幫助。

這裏是我的功能代碼：

create or replace function janek.split_table (table_schema text, table_name text, column_name text, separator text) 
    RETURNS setof $1.$2 AS 
    -- don't know how to pass table_schema.table_name as variable - need help here, this line doesn't work, but after manually entering schema and table name it works 
$BODY$ 
DECLARE 

    -- string variable containing sql queries to execute 
    execute_query text; 

    -- string variable where column list of target table will be assigned, except for one column - column_name 
    column_list text; 

    -- variable where maximal count of separator will be assigned 
    separator_count integer; 

BEGIN 

    -- assigning variable column_list 

    -- first error here, that I don't understand - when I execute this select statement manually it returns one column containing one string: 
    -- ERROR: query "SELECT 'array_to_string(array(select column_name::text 
    -- from information_schema.columns 
    -- where table_name = '''||table_name||''' 
    -- and table_schema = '''||table_schema||''' 
    -- and column_name <> '''||column_name||''' 
    -- ), ', ');'" returned 2 columns 
    -- CONTEXT: PL/pgSQL function "split_table" line 15 at assignment 
    execute_query := 'SELECT array_to_string(array(select column_name::text 
    from information_schema.columns 
    where table_name = '''||table_name||''' 
    and table_schema = '''||table_schema||''' 
    and column name <> '''||column_name||''' 
    ), ', ');'; 
    execute execute_query into column_list; 

    -- assigning variable separator_count - code of function janek.cointinstring below, it returns number of occurences of separator in string 
    execute_query := 'select max(janek.countinstring('''||column_name||''', '''||separator||''')) from '||table_schema||'.'||table_name||''''; 
    execute execute_query into separator_count; 

    -- redefining table we're transforming - splitting column_name to an array on delimiter + deleting spaces. If you guys know a way to remove spaces only 'touching' the separator it would be better: 
    execute_query := 'with t as (
    select '||column_list||', 
    case when position('''||separator||''' in '||table_name||') > 0 then 
    regexp_split_to_array(replace('||column_name||', '' '', ''''), '''||separator||''') 
    elsif length('||table_name||') > 0 then array['||column_name||'] 
    else null 
    as arr 
    from '||table_schema||'.'||table_name||')'; 

    -- main loop, iterates on table in order to transform column 'arr' to separate strings instead of array using union all: 
    for i in 1..separator_count loop 
     execute_query := execute_query||' 
     select 
     '||column_list||', arr['||separator||'] as '||column_name||' 
     from t 
     where arr['||separator||'] is not null 
     union all'; 
    end loop; 

    -- removing last union all 
    execute_query := substr(execute_query, 1, length(execute_query) - 9); 

    -- executing the main query 
    RETURN QUERY EXECUTE execute_query; 

END $BODY$ 
    LANGUAGE plpgsql VOLATILE 
    COST 100; 
ALTER FUNCTION janek.split_table (table_schema text, table_name text, column_name text, separator text) 
    OWNER TO jsiekierski; 
GRANT EXECUTE ON FUNCTION janek.split_table (table_schema text, table_name text, column_name text, separator text) TO jsiekierski; 
GRANT EXECUTE ON FUNCTION janek.split_table (table_schema text, table_name text, column_name text, separator text) TO wsd_users; 


-- janek.countinstring function code: 
CREATE OR REPLACE FUNCTION janek.countinstring(text, text) 
    RETURNS integer AS 
$BODY$ 
SELECT(Length($1) - Length(REPLACE($1, $2, '')))/Length($2) ; 
$BODY$ 
    LANGUAGE sql IMMUTABLE 
    COST 100; 
ALTER FUNCTION janek.countinstring(text, text) 
    OWNER TO jsiekierski; 

Answer 1

有一個regexp_split_to_table()，很像regexp_split_to_array()您正在使用：

http://www.postgresql.org/docs/current/static/functions-string.html

select regexp_split_to_table('hello world', E'\\s+'); 

但你的問題聞起來像你正試圖重塑COPY ... DELIMITER ...命令：

http://www.postgresql.org/docs/current/static/sql-copy.html

PS - 提示：使用雙美元作爲喜歡你寫的那些功能薈萃字符串分隔符：select $sql$foo, bar, baz$sql$;