使用连接更新 Redshift 表

Question

I have 3 table t1, t2 and t3.
t1 has 2 column-> id1, val1

t2 -> id2, val2  
t3 -> id3, val3  
If id1=id2 and id2 = id3 

then I need to update val1 ad val3. But I have repeating id1 and each should have same val3

I am using

update  t1
inner join  t2 on t1.id1 = t2.id2
inner join  t3 on t2.id2 = t3.id3
set t1.val1 =  t3.val3
;

But not able to do this.

Answer 1

The correct syntax is:

UPDATE table_name SET column = { expression | DEFAULT } [,...]
[ FROM fromlist ]
[ WHERE condition ]

So your UPDATE statement should look as follows:

update t1 set val1 = val3
from t2 inner join t3 on t2.id2 = t3.id3
where t1.id1 = t2.id2
;

See the Redshift documentation and their comprehensive UPDATE examples .

Answer 2

I needed to get values from the other table, t2.val3, on Redshift. I used the following

update t1 
set val1 = t2.val3
from t2 join t1 t on t.id = t2.id;

I have to re-name t1. Otherwise Redshift complains.

Answer 3

Per @jie solution, but expanded a bit and w/o the distraction of a table alias as a table name (heh heh):

update
  my_counts
set 
  my_count_delta = t1.my_count - t2.my_count
from
  my_counts t1 join
  my_counts t2 on
    t1.group_id = t2.group_id and 
    t2.count_dt = ( t1.count_dt - interval '1 day' )
where
  t1.count_error is null

A few notes:

• my_count is a running total
• my_count_delta is the change from the previous day entry (-/+)
• this solves the issue where running total exists, but summarial columnar data for the delta needs to be added
• :heart: how pg style sql makes date add/subtraction so simple: ( t1.count_dt - interval '1 day' )
• as an avid lover of LEFT JOIN i was confounded that this would not run with a left join...the error message was very clear requiring "balanced join" or was it "even join" ...so i went back to the @jie version of JOIN and found that the query was incredibly fast .