Question

我在确定哪些日期超过1％的状态错误时遇到了一些麻烦：

 path   | text                     |
 ip     | inet                     |
 method | text                     |
 status | text                     |
 time   | timestamp with time zone | default now()
 id     | integer                  | not null default nextval('log_id_seq'::regclass)
Indexes:
    "log_pkey" PRIMARY KEY, btree (id)

重要的部分是200 OK的状态或错误和时间，这是日期，其余的对于这个特殊的问题有点无用我相信

这是我到目前为止的代码：

def heavy_error_days():
    db = psycopg2.connect("dbname=news")
    c = db.cursor()
    c.execute("select date(log.time), errors\
     from log, (select count(status)::numeric/(select count(status)\
     from log)from log where status <> '200 OK'\
     and date(log.time) = date(log.time)) as errors\
     group by date, errors")
    print c.fetchone()

我相信我需要做的是将误差乘以100并将它们分成各自的日期，但我不知道如何写它。

如果有人能帮助我，我将非常感激。

错误

  File "news.py", line 33
    c.execute("with a as (select distinct (sum(case when status <> '200 OK' then 1 else 0 end) over w * 100) / count(1) over w perc, "time"::date d from log window w as (partition by "time"::date))select * from a where perc > 1")
                                                                                                                                         ^
SyntaxError: invalid syntax

我注意到时间是双引号，这可能是问题因此我将其更改为单引号并出现此错误：

Traceback (most recent call last):
  File "news.py", line 41, in <module>
    heavy_error_days()
  File "news.py", line 33, in heavy_error_days
    c.execute("with a as (select distinct (sum(case when status <> '200 OK' then 1 else 0 end) over w * 100) / count(1) over w perc, 'time'::date d from log window w as (partition by 'time'::date))select * from a where perc > 1")
psycopg2.DataError: invalid input syntax for type date: "time"
LINE 1: ...else 0 end) over w * 100) / count(1) over w perc, 'time'::da...
                                                             ^

Answer 1

这个查询应该可以解决问题：

with a as (
  select distinct (sum(case when status <> '200 OK' then 1 else 0 end) over w * 100) / count(1) over w perc, "time"::date d
  from log
  window w as (partition by "time"::date)
)
select *
from a
where perc > 1

基于例子：

构建

t=# create table log("time" timestamptz, status text);
CREATE TABLE
t=# insert into log values (now(),'200 OK'),(now(),'200 OK'),(now(),'ERR'),(now()-'2 days'::interval,'200 OK');
INSERT 0 4

查询：

t=# with a as (
  select distinct (sum(case when status <> '200 OK' then 1 else 0 end) over w * 100) / count(1) over w perc, "time"::date d
  from log
  window w as (partition by "time"::date)
)
select *
from a
where perc >= 0
;
 perc |     d
------+------------
    0 | 2017-05-17
   33 | 2017-05-19
(2 rows)

Python psql拆分成各自的列

1 个答案: