I'm migrating some MySQL code to Postgres and having a heck of a time converting the following line:
GROUP_CONCAT(
DISTINCT id,':',foo,':',bar ORDER BY id
) as id,
This results in a comma separated list of strings like:
id:foo:bar,id2:foo2:bar2
The 'DISTINCT' is there to avoid duplicates.
I've read the equivalent of GROUP_CONCAT in Postgres is string_agg, but I can't figure out how to make it work the same way.
Edit: I may have almost answered my own question. I came up with the solution of using CONCAT. The problem with this is that now sorting is done by the concatenated string rather than by the id:
string_agg(DISTINCT CONCAT(
id::text, ':', foo, ':', bar, ':'
), ',') as id
If I try to add 'ORDER BY id' I get an error.
You can do something like below.
select
string_agg(DISTINCT CONCAT(
id::text, ':', foo, ':', bar, ':'
), ',') as id from (select * from table t order by id) as al
Related
I have a table with a varchar column that represents a path. I want to search for rows that have a path that follow a pattern like name.name[*] where name can be anything. I am looking for repeated strings contained anywhere in the path column that are separated by a period and have a square bracket after them.
This seems to call for Regexp, so through python I have something like https://regex101.com/r/apS20a/4
However, trying to implement this with MySQL Regexp is not working. I have been able to translate the shorthand into REGEXP '([A-Za-z_]+).(\1[[0-9]+])', but it seems that MySql Regex does not support capture groups. Is there a way to accomplish what I am trying to do with mysql regexp? Thank you
I don't think that MySQL supports capture groups. But if you only have one example of .name[ in the string between the first . and the first [, you can hack your way around it. This is not a general solution, just a specific approach in this case.
You can get the name with:
select substring_index(substring_index(url, '[', 1), '.', -1) as name
And then incorporate this into a regular expression:
select t.*
from (select t.*,
substring_index(substring_index(url, '[', 1), '.', -1) as name
from t
) t
where url like concat('%', name, '.', name, '[%');
This just uses like instead of regexp, because [ and . are regular expression wildcards. Of course, this assumes that name does not have _ or %.
EDIT:
Here is a method that actually identifies when this occurs -- and works even if there are multiple patterns.
The idea is to construct the regular expression based on what happens between the . and [ -- and then to apply it. Delightfully self-referential:
select t.*,
(url regexp regex)
from (select t.*,
substr(regexp_replace(url, '[^.]*[.]([^\\[]*)\\[[^.]*', '|$1[.]$1\\\\['), 2) as regex
from (select 'abcde.de[12345.345[ABC' as url union all
select 'abcdefdef[[[[..123.124['
) t
) t;
Here is the above in a db<>fiddle.
Given data like:
URL
some_url.com
some_url.com
some_url.co.uk
some_other_url.com
some_other_url.co.uk
some_other_url.co.uk
some_other_url.org
is there a way to construct a query that will result in;
some_url 3
some_other_url 4
Currently I'm either using a standard group by url or I query the aggregations one by one using LIKE
Is there a way to do this in one query? (using mysql currently, but will be moving this data over to postgresql)
Would it be better practice to add a column to reflect this grouping (at insert time)? (this feels redundant but would be best performing I guess)
EDIT:
data can contain www and non-www as well as http, https. Also I'll have to do similar thing on other columns that contain (free) text values.
This is ANSI SQL compliant and should probably work with both MySQL and Postgresql:
select url, count(*)
from
(
select substring(url from 1 for position('.' in url) -1) as url
from tablename
) dt
group by url
Using position() to find the first . character. Do substring() and finally GROUP BY the result.
use SUBSTRING_INDEX in mysql which help you substring from a string before a specified number of occurrences of the delimiter.
select count(*) as cnt, SUBSTRING_INDEX(c,'.',1) as val from cte
group by SUBSTRING_INDEX(c,'.',1)
Since the values can have http, https and www, and may be query string too, you will have to clean all such values first before grouping it. Took the reference from here and modified it to match your requirement.
SELECT url,
SUBSTRING_INDEX(
SUBSTRING_INDEX(
SUBSTRING_INDEX(
SUBSTRING_INDEX(
SUBSTRING_INDEX(
SUBSTRING_INDEX(url, '/', 3),
'://', -1),
'/', 1),
'?', 1),
'www.', -1),
'.', 1) AS domain,
COUNT(1)
FROM tblname
GROUP BY domain;
This works in Postgesql:
select split_part(url,'.',1) g,count(*)
from url_table
group by g
order by g;
Best regards,
Bjarni
I have a data that have this result:
I'm using the query
select substring_index(descriptn, ' ', -1) from table1
and I get my result along with the dash for example "JACQ-ARMIE"
I want to get only JACQ and ABBY. Can you give me hints on how to get the name? Does this also apply for substring_index?
Just use substring_index() again:
select substring_index(substring_index(descriptn, ' ', -1), '-', 1)
from table1;
I'm new to Presto and looking to get the same functionality as the group_concat function in MySQL. Are the following two equivalent? If not, any suggestions for how I can recreate the group_concat functionality in Presto?
MySQL:
select
a,
group_concat(b separator ',')
from table
group by a
Presto:
select
a,
array_join(array_agg(b), ',')
from table
group by a
(Found this as a suggested Presto workaround here when searching group_concat functionality.)
Try using this in place of group_concat in Presto ::
select
a,
array_join(array_agg(b), ',')
from table
group by a
Also, if you're looking for unique values only – an equivalent to group_concat(distinct ... separator ', ') – try this:
array_join(array_distinct(array_agg(...)), ', ')
There's no function as of this answer, though the feature has been requested.
The closest equivalent is mentioned in your question.
WITH tmp AS (
SELECT 'hey' AS str1
UNION ALL
SELECT ' there'
)
SELECT array_join(array_agg(str1), ',', '') AS joined
FROM tmp
I have a table which i am using to query and getting its one column which matches regular expression which is (\/.+\/\?).
Content of the resulted column is like:
/Anything here/?
Example output:
\abc\cdf\?....
\ab\?....
\abc\cdf\?....
\sb\?....
where '....' can be anything
Desired result i want is unique values before \? such that rows with duplicate regexp matched content are shown once only like here (\abc\cdf\?.... showing twice instead of onece)
\abc\cdf\?....
\ab\?....
\sb\?....
OR
\abc\cdf\?
\ab\?
\sb\?
I have looked very much but couldn't find anything there is regexp_substr in oracle but that is not working in SQL.
Please if someone could help me with the sql query that would be awesome.
If you want everything before the last \, then you can use substring_index() and some string manipulation:
select substring_index(col, '\\',
length(col) - length(replace(col, '\\', ''))
) as firstpart,
count(*)
from table t
group by substring_index(col, '\\',
length(col) - length(replace(col, '\\', ''))
);