Let's say we have following table.
UserId | Message
-------|-------------
1 | Hi, have a nice day
2 | Hi, I had a nice day
I need to have all { Hi,-have-a-nice-day-I-had } words separately.
Is there any way to do that ? What if I want to export words from whole database tables ?
Similar results would be also good.
try this:In Sql server 2005 or above
create table yourtable(RowID int, Layout varchar(200))
INSERT yourtable VALUES (1,'hello,world,welcome,to,tsql')
INSERT yourtable VALUES (2,'welcome,to,stackoverflow')
;WITH SplitSting AS
(
SELECT
RowID,LEFT(Layout,CHARINDEX(',',Layout)-1) AS Part
,RIGHT(Layout,LEN(Layout)-CHARINDEX(',',Layout)) AS Remainder
FROM YourTable
WHERE Layout IS NOT NULL AND CHARINDEX(',',Layout)>0
UNION ALL
SELECT
RowID,LEFT(Remainder,CHARINDEX(',',Remainder)-1)
,RIGHT(Remainder,LEN(Remainder)-CHARINDEX(',',Remainder))
FROM SplitSting
WHERE Remainder IS NOT NULL AND CHARINDEX(',',Remainder)>0
UNION ALL
SELECT
RowID,Remainder,null
FROM SplitSting
WHERE Remainder IS NOT NULL AND CHARINDEX(',',Remainder)=0
)
SELECT part FROM SplitSting ORDER BY RowID
SQLFIDDLE DEMO
Well, ok, here it goes.
In SQL Server you can use this...
SELECT word = d.value('.', 'nvarchar(max)')
FROM
(SELECT xmlWords = CAST(
'<a><i>' + replace([Message], ' ', '</i><i>') + '</i></a>' AS xml)
FROM MyMessageTbl) T(c)
CROSS APPLY c.nodes('/a/i') U(d)
And I hope that for MySQL you can use the same thing, using XML support - ExtractValue() etc.
EDIT: explanation
- replace([Message], ' ', '</i><i>') replaces e.g. 'my word' with 'my</i><i>word'
- then I add the beginning and the end of xml -> '<a><i>my</i><i>word</i></a>', so I have a valid xml... and cast it to xml type to be able to do something with it
- I select from that xml and shred xml nodes '/a/i' it to rows using CROSS APPLY c.nodes('/a/i');
alias rows using U(d), so one 'i' maps to column d (e.g. 'my')
- d.value('.', 'nvarchar(max)') extracts node content and casts it to character type
Related
I am trying to replace substrings within one text column in my table using a reference table.
To my knowledge, the replace(column, string1,string2) function will only work with strings as the second and third input.
Here is a visual of what I am trying to do. To be clear, the reference table I need to use is much larger - otherwise, I would use four replace functions.
EDIT: Thank you to everyone who has pointed out how bad this data model is built. Though I am not an expert on building efficient data models, I do know this one is built terribly. However, the structure of this model is completely out of my control. Apologies for not mentioning that from the get-go.
table1
Farms
Animals
Farm1
Cow, Pig
Farm2
Dog, Cow, Cat
Farm3
Dog
referenceTable
refColumn1
refColumn2
Cow
Moo
Pig
Oink
Dog
Bark
Cat
Meow
And here is what I would like the result column to be..
table1
Farms
Animals
Farm1
Moo, Oink
Farm2
Bark, Moo, Meow
Farm3
Bark
First question on stackoverflow so apologies if I missed anything.
Any help is appreciated! Thank you!
To loop over comma (or ', ' in this case) separated values, you can use a double substring_index and a join against a sequence table (where the sequence is <= the number of joined values in a given row, as determined with char_length/replace):
select t1.Farms, group_concat(rt.refColumn2 order by which.n separator ', ') Animals
from table1 t1
join (select 1 n union select 2 union select 3) which
on ((char_length(t1.Animals)-char_length(replace(t1.Animals,', ','')))/char_length(', '))+1 >= which.n
join referenceTable rt on rt.refColumn1=substring_index(substring_index(t1.Animals,', ',which.n),', ',-1)
group by t1.Farms
Here I use an ad hoc sequence table of 1 through 3, assuming no row will have more than 3 animals; expand as necessary or alternatively use a cte.
You have a really lousy data model and you should fix it. You should not be storing multiple values in a string column. Each value pair should be on its own row.
Let me assume that someone else created these tables and you have no choice. If that is the case, MySQL has a solution. I think I would suggest:
select t1.*, -- or whatever columns you want
(select group_concat(rt.refColumn2
order by find_in_set(rt.refColumn1, replace(t1.animals, ', ', ','))
separator ', '
)
from referenceTable rt
where find_in_set(rt.refColumn1, replace(t1.animals, ', ', ',')) > 0
)
from table1 t1
I'm more fluent in Sql Server than MySql, having got a solution working in Sql Server the real challenge was converting to a working MySql version!
See if this meets your needs. It works for your sample data, you may of course need to tweak if it doesn't fully represent your real world data.
with w as (
select *, case when animals like '%' || refcol1 || '%' then locate(refcol1,animals) end pos
from t1
join lateral (select * from t2)t2 on 1=1
)
select farms, group_concat(refcol2 order by pos separator ',') as Animals
from w
where pos>0
group by farms
order by farms
Working DB<>Fiddle
I have a varchar(255) field within a source table and the following contents:
50339 My great example
2020002 Next ID but different title
202020 Here we go
Now I am processing the data and do an insert select query on it. From this field I would need the INT number at the beginning of the field. IT IS followed by 2 spaces and a text with var length, this text is what I need as well but for another field. In General I want to to put text and ID in two fields which are now in one.
I tried to grab it like this:
SELECT STATUS REGEXP '^(/d{6,8}) ' FROM products_test WHERE STATUS is not null
But then I learned that in MySQL 5.x there are no regexp within the SELECT statement.
How could I seperate those values within a single select statment, so I can use it in my INSERT SELECT?
From the correct solution of user slaakso, resulted another related problem since somtimes the STATUS field is empty which then results in only one insert, but in case there is a value I split it into two fields. So the count does not match.
My case statement with his solution somehow contains a syntax problem:
CASE STATUS WHEN ''
THEN(
NULL,
NULL
)
ELSE(
cast(STATUS as unsigned),
substring(STATUS, locate(' ', STATUS)+3)
)
END
You can do following. Note that you need to treat the columns separately:
select
if(ifnull(status, '')!='', cast(status as unsigned), null),
if(ifnull(status, '')!='', substring(status, locate(' ', status)+2), null)
from products_test;
See db-fiddle
I have a database with TV Guide data, and in my description field (VARCHAR) sometimes i have a '|' where behind it is the rating. I used to check this in php, before converting it all to XML, but i would like to do this in SQL.
So if i have this string:
This is the description | rating pg-13
Then i want to keep the
This is the description
but if there is no '|' i want the whole string.
I tried using substring, but can't get it to work.
My query now is:
SELECT *, SUBSTRING(`long_description`, 1, POSITION('|' IN `long_description`)) FROM `programs` WHERE station_id = 1
this works only one way - this gives me the string before the '|' but if there is no '|' it gives an empty column.
Based on the use of backticks, you might be using MySQL. If so, substring_index() does exactly what you want:
select substring_index(long_description, '|', 1)
How about this:
SELECT
*,
IF(long_description LIKE '%|%',
SUBSTRING(`long_description`,
1,
POSITION('|' IN `long_description`)),
long_description)
FROM
`programs`
WHERE
station_id = 1
The IF clause basically just checks if you have a | in the field and applies your routine when this is true. Else it will simply return the complete long_description value.
I am working on the sql query in which I want to sort the string on the basis of numbers.
I have one column (Column Name is Name) table in which there are multiple fields. On using ORDER BY NAME, it prints in the following way:
hello_world
hello_world10
hello_world11
hello_world12
hello_world13
hello_world14
hello_world15
hello_world4
hello_world5
For the above query, I have used ORDER BY NAME; but it doesn't seem to print on the basis of numbers.
Problem Statement:
I am wondering what sql query I need to write or what changes I need to make in my sql query above so that it prints everything on the basis of numbers, the o/p should be this:
hello_world
hello_world4
hello_world5
hello_world10
hello_world11
hello_world12
hello_world13
hello_world14
hello_world15
you want a numeric ordering, then you need to create a numeric value to order on.
currently you have strings.
if the pattern is true, then you can use a combination of string manipulation to trim off the first characters, which should leave only numbers, then use TO_NUMBER() to convert for the ordering
something like
select name
from mytable
order by to_number( replace( name, 'hello_world','' ))
I think the simplest solution for this particular case (where all the values have the same prefix) is:
order by length(name), name
Try this:
SELECT name,
CASE WHEN REGEXP_INSTR(name, '[0-9]') = 0 THEN 0
ELSE CAST(SUBSTR(name, REGEXP_INSTR(name, '[0-9]')) AS INT)
END AS progressive
FROM my_table
ORDER BY progressive;
we can order it using replace and cast methods.
I tried the following query
select Name, cast(REPLACE(Name, 'hello_world', '') as UNSIGNED ) as repl from Users order by repl;
To generage sample data
CREATE TABLE Users (
Name varchar(255) NOT NULL
);
insert into Users(Name) values
('hello_world'),
('hello_world4'),
('hello_world5'),
('hello_world10'),
('hello_world11'),
('hello_world12'),
('hello_world13'),
('hello_world14'),
('hello_world15')
;
EDIT
query without replaced column,
select City from Persons order by cast(REPLACE(City, 'hello_world', '') as UNSIGNED );
Though the question is about mysql.
I tried in sql server.
create table #t1 (id varchar(100));
insert into #t1 (id) values ('Pq1'),('pq3'),('pq2')
select * from #t
order by
CAST(SUBSTRING(id + '0', PATINDEX('%[0-9]%', id + '0'), LEN(id + '0')) AS INT)
I have a table called SerialNos with the following columns:
Id, Parent_Id, SerialNo
There are different SerialNo on a Parent_Id
Like:
Id Parent_Id SerialNo
1 16 abc
2 16 def
3 23 hij
4 23 klm
5 23 nop
I just want to retrieve comma separated SerialNos. for particular Parent_Id
e.g. If Parent_Id is passed 16,then O/p should be : 'abc,def'
and if Parent_Id is passed 23,then O/p should be : 'hij,klm,nop'
here is a nice hack:
DECLARE #csv varchar(1000)
SELECT #csv = COALESCE(#csv+',','') + SerialNo
FROM SerialNos
WHERE Parent_Id = 23
SELECT #csv
Depending upon the database server you can use group_concat. For example if you are using mysql group_concat with a separator of "," will give you the result you are after.
http://dev.mysql.com/doc/refman/5.0/en/group-by-functions.html#function_group-concat
If you are using a different database server check its documentation it is likely that it will have something similar. Please also keep in mind that some databases don't have this function.
Sorry just saw that your post was tagged sql-server2008. If sql server 2008 does not have group_concat or something similar try this link
http://blog.shlomoid.com/2008/11/emulating-mysqls-groupconcat-function.html
Sorry yet another edit, this will help.
http://groupconcat.codeplex.com/
for sql-server 2008 you can use the following query(with your table name):
select distinct STUFF(ISNULL((SELECT ', ' + x.SerialNo
FROM TableA x
WHERE x.Parent_Id= t.Parent_Id
GROUP BY x.SerialNo
FOR XML PATH (''), TYPE).value('.','VARCHAR(max)'), ''), 1, 2, '')
from TableA t
where Parent_Id = 16
Well i could advice my blog post here, but unfortunately its not in English :).
Here is something similar, just without CLR.
There i am aware of at least 4 solutions:
CLR grouping function (well, code is in english)
Can use cursors (not very fast)
Can use something similar to cursors
Can use XML
So one solution (not the fastest, but easy to write):
Create Function fn_MyFunction
(
#Parent_Id int
)
Returns NVarChar(1000)
As
Begin
Declare #txt NVarChar(1000)
SELECT #txt = COALESCE(#txt + ';' + SerialNo, SerialNo)
FROM dbo.Table Where Parent_Id = #Parent_Id
Return #txt
End