PostgreSQL：将一个非常大的数字的十六进制字符串转换为数字

Question

我正在尝试转换一个非常大的数字的十六进制字符串 到 NUMERIC 列

CREATE OR REPLACE FUNCTION hex_to_int(hexval varchar) RETURNS NUMERIC AS $$
DECLARE
   result  NUMERIC;
BEGIN
 EXECUTE 'SELECT x''' || hexval || '''::NUMERIC(40,0)' INTO result;
 RETURN result;
END;
$$
LANGUAGE 'plpgsql' IMMUTABLE STRICT;

我正在尝试这样做：

select hex_to_int(tx.value) from internal_transaction tx

我得到的错误是：

[42846] ERROR: cannot cast type bit to numeric Where: PL/pgSQL function hex_to_int(character varying) line 5 at EXECUTE statement

Answer 1

又一个转换函数。这个想法是减少循环中的步骤数，从而减少算术运算的数量。

create or replace function hex_to_numeric(str text)
returns numeric language plpgsql immutable strict as $$
declare
    i int;
    n int = length(str)/ 8;
    res dec = 0;
begin
    str := lpad($1, (n+ 1)* 8, '0');
    for i in 0..n loop
        if i > 0 then 
            res:= res * 4294967296; 
        end if;
        res:= res + concat('x', substr(str, i* 8+ 1, 8))::bit(32)::bigint::dec;
    end loop;
    return res;
end $$;

一些测试：

select hex, hex_to_numeric(hex) 
from   (
   values ('ff'::text),
        ('7fffffff'),
        ('80000000'),
        ('deadbeef'),
        ('7fffffffffffffff'),
        ('8000000000000000'),
        ('ffffffffffffffff'),
        ('ffffffffffffffff123'),
        ('4540a085e7334d6494dd6a7378c579f6')
   ) t(hex);
   
               hex                |             hex_to_numeric             
----------------------------------+----------------------------------------
 ff                               |                                    255
 7fffffff                         |                             2147483647
 80000000                         |                             2147483648
 deadbeef                         |                             3735928559
 7fffffffffffffff                 |                    9223372036854775807
 8000000000000000                 |                    9223372036854775808
 ffffffffffffffff                 |                   18446744073709551615
 ffffffffffffffff123              |                75557863725914323415331
 4540a085e7334d6494dd6a7378c579f6 | 92052294502540680826164862654582454774
(9 rows)

该函数比其他答案中的 plpgsql 函数快 5-10 倍（取决于十六进制值的长度）。

Answer 2

这是一种蛮力，一点也不防弹：

CREATE OR REPLACE FUNCTION hex_to_int(hexval varchar) RETURNS numeric AS $$
DECLARE
  result  NUMERIC;
  i integer;
  len integer;
  hexchar varchar;
BEGIN

  result := 0;
  len := length(hexval);

  for i in 1..len loop
    hexchar := substr(hexval, len - i + 1, 1);
    result := result + round(16 ^ (i - 1)::dec * case
      when hexchar between '0' and '9' then cast (hexchar as int)
      when upper (hexchar) between 'A' and 'F' then ascii(upper(hexchar)) - 55
    end);
  end loop;

 RETURN result;
END;
$$
LANGUAGE 'plpgsql' IMMUTABLE STRICT;

select hex_to_int('12AE34F');  -- returns 19587919

或者，如果您安装了 PL/Perl，您可以让它完成繁重的工作：

CREATE OR REPLACE FUNCTION hex_to_int_perl(varchar)
  RETURNS numeric AS
$BODY$
  my ($hex) = @_;
  return sprintf "%d", hex($hex);
$BODY$
  LANGUAGE plperl VOLATILE
  COST 100;

select hex_to_int_perl('12AE34F');  -- returns 19587919

我认为非 Perl 不适用于负数，而且我很确定如果输入非十六进制值，两者都会给你带来不好的结果，但这些场景很容易陷入陷阱和句柄，取决于你想让函数做什么。

Answer 3

我的用例是将十六进制 SHA-1 哈希值转换为整数。如果不注意整个函数的数值精度，这个用例很容易暴露缺点；但当然，具有更大“非常大数字”的用例很容易识别。

给定一些 SHA-1 哈希，第一个接受的答案的解决方案产生以下结果：

SELECT hex_to_int('356e90d2a2d414ba8757ec2ab91f2f19c481d4c3');
-- returns 305042208670409000000000000000000000000000000000

SELECT hex_to_int('aaa9f7193cc8efe7e98145b0f8d9ae5f1712c25b');
-- returns 974318782301086000000000000000000000000000000000

当然，每个结果右填充了这么多零是用例精度不足的标志。

这是我最终获得所需精度的地方：

CREATE OR REPLACE FUNCTION hex_to_int(hexVal varchar) RETURNS numeric(1000) AS $$
DECLARE
    intVal numeric(1000) := 0;
    hexLength integer;
    i integer;
    hexDigit varchar;
BEGIN
    hexLength := length(hexVal);

    FOR i IN 1..hexLength
    LOOP
        hexDigit := substr(hexVal, hexLength - i + 1, 1);
        intVal :=
            intVal +
            CASE
                WHEN hexDigit BETWEEN '0' AND '9' THEN CAST(hexDigit AS numeric(1000))
                WHEN upper(hexDigit) BETWEEN 'A' AND 'F' THEN CAST(ascii(upper(hexDigit)) - 55 AS numeric(1000))
            END *
            CAST(16 AS numeric(1000)) ^ CAST(i - 1 AS numeric(1000));
    END LOOP;

    RETURN intVal;
END;
$$
LANGUAGE 'plpgsql' IMMUTABLE STRICT;


SELECT hex_to_int('356e90d2a2d414ba8757ec2ab91f2f19c481d4c3');
-- returns 305042208670409212880202819376501392142011323587

SELECT hex_to_int('aaa9f7193cc8efe7e98145b0f8d9ae5f1712c25b');
-- returns 974318782301085717223606317572453925850501530203