How to find max value to find partially duplicate rows

[chintan85]

I have a table1

id col1 col2
1 NC_001A>T 5
2 NC_001C>G 4
3 NC_001G>C 3
4 NC_001_98_G>C 1

Now, I want to have max col2 value for partially duplicate col1 value

for example if I take partial value for col 1 values NC_001A>T, NC_001C>G and NC_OO1G>C, I can write following sql query

 select col1, MAX (col2) from table1 where col1 = ('NC_001')group by col1

Result:
id col1 col2
1 NC_001A>T 5



But I am not sure how to find max value for all similar partically duplicate rows i.e NC_002, NC_003 and so on in the table

Thanking in advance.

[Rabbit]

First off, you have to define "partially duplicate" since that can mean a hundred different things.

[chintan85]

Partially duplicate values are found in similar pattern in col1.

for example consider
NC_001A>T, NC_001C>G are one pair partially similar
NC_002A>T, NC_002C_G are another pair partially similar

they all differ by having expression such as C>G or G>T

[zmbd]

ChinTan
I think that you're going to have to start by defining what is the same about the data and what makes it the largest.

-z

[Rabbit]

That still doesn't explain exactly what you mean by "partially duplicate". In your first post, is that fourth item (NC_001_98_G>C) a partial duplicate? Are you defining partial duplicate as the first 6 characters? Which would mean the fourth item is a duplicate. Or do you mean that if you strip off the last 3 characters, if the remainder matches, then it's a duplicate? In which case the fourth item is not a duplicate? Or do you have some other definition of duplicate?

[chintan85]

Sorry I was not clear but please exclude row 4 NC_001_98_G>C 1.
Let just concentrate on values which are same when stripping of last three character (i.e A>T or it could be c>G or G>T etc..)

"Or do you mean that if you strip off the last 3 characters, if the remainder matches, then it's a duplicate?"

Yes, I want to write a sql query that matches these values and find max 'col2' value for them.

Is that clear?

[Rabbit]

What you can do then is use a combination of the left function with the len function to strip off the last 3 characters. Use that in place of your col1 in your previous query and that will give you what you want.

[chintan85]

select left(col1,LEN(col1) - 3), MAX (col2) from table1 group by col1

The above the query results in:
NC_001 5

which is max from
1 NC_001A>T 5
2 NC_001C>G 4
3 NC_001G>C 3


now how can I modify query to produce result
NC_001A>T 5

which is max from

[chintan85]

Oh I am sorry the query failed. Can you correct me what I am doing wrong

select left(col1,LEN(col1) - 3), MAX (col2) from table1 group by col1

I have tried to use where clause but it failed

[Rabbit]

There's no where clause in your query.

You can get the original value by joining it back to the table on the partial key and value.

[chintan85]

can you modify query to give correct result

select left(col1,LEN(col1) - 3) as , MAX (col2) from table1 group by col1

[Rabbit]

You need to group by your new calculated field. You also need to give your calculated field an alias.

[chintan85]

I found a smarter way to do this. The below query fetches the row from col1 value where no other row exists with the same col1 value and a greater col2 values for that.

SELECT *
FROM table1 as t1
LEFT OUTER JOIN table1 AS t2
ON (left(t1.col1,LEN(t1.col1) - 3)= left(t2.col1,LEN(t2.col1) - 3) AND t1.col2 < t2.col2)
WHERE t2.col1 IS Null

[ck9663]

Make sure t2.col1 does not allow NULL. Also, if this will be executed against a large table, use NOT EXISTS instead.

Good Luck!!!


~~ CK