Why are queries these not equivalent? (correlated subquery vs. group by) - sql

Question

Why are these two SQL queries not equivalent? One uses a correlated subquery, the other uses group by. The first produces a little over 51000 rows from my database, the second nearly 66000. In both cases, I am simply trying to return all the parts meeting the stated condition, current revision only. A comparison of the output files shows that method #1 (oracle_test1.txt) fails to return quite a few values. Based on that, I can only assume that method #2 is correct. I have some code that has used method #1 for a long time, but it appears I will have to change it. My reasoning concerning the correlated subquery was that as the inner select is comparing the columns in the self join, it will find the max vaule for the prev value for all matches; then return that max prev value for use in the outer query. I designed that query long ago before becoming familiar with the use of group by. Any insights would be appreciated.
Query #1
select pobj_name, prev
from pfmc_part
where pmodel in ('PN-DWG', 'NO-DWG') and pstatus = 'RELEASED'
and prev = (select max(prev) from pfmc_part a where a.pobj_name = pfmc_part.pobj_name)
order by pobj_name, prev"
Query #2
select pobj_name, max(prev) prev
from pfmc_part
where pmodel in ('PN-DWG', 'NO-DWG') and pstatus = 'RELEASED'
group by pobj_name
order by pobj_name, prev"
Sample output:
Query #2 Query #1
P538512 B P538512 B
P538513 A P538513 A
P538514 C P538514 C
P538520 B
P538522 B P538522 B
P538525 A P538525 A
P538531 C P538531 C
P538533 A P538533 A
P538538 B
P538541 B
P538542 B
P538553 A P538553 A
P538569 A P538569 A

Answer 1

Query 1 is returning each of the max ids and then those that have a pmodel of the type specified within your where clause.
Whereas query 2 is selecting all items with a pmodel of the type specified in your where clause and each of the max ids of that.
You may have data which isn't the max id which satisfies your where clause in query 2 which is why it's being omitted in query 1

Answer 2

There are two differences and the rest of the answers focus on one. The "easy" difference is that the max() in the group by is affected by the filter clause. The max() in the other query has no filter, and so it might return no rows (when max(prev) is on a row otherwise filtered out by the where conditions).
In addition, the where version of the query might return duplicate rows when there are multiple rows with the same value of max(prev) for a given pobj_name. The group by will never return duplicate rows.

Answer 3

this query
select pobj_name, prev
from pfmc_part
where pmodel in ('PN-DWG', 'NO-DWG') and pstatus = 'RELEASED'
and prev = (select max(prev) from pfmc_part a where a.pobj_name = pfmc_part.pobj_name)
order by pobj_name, prev"
has a where clause declaration causing it to return less rows -- specifically, only rows where prev = (subquery). that and prev makes it entirely different, and also assigns the value into prev in the first line
if you wanted them to be the more similar, you'd need to modify it like so
select pobj_name, prev, maxes.max
from pfmc_part
JOIN (select max(prev) as max from pfmc_part a where a.pobj_name = pfmc_part.pobj_name) maxes
where pmodel in ('PN-DWG', 'NO-DWG') and pstatus = 'RELEASED'
order by pobj_name, prev"

Answer 4

In query 1 you are ONLY selecting the rows whose prev field is equal to the max(prev) and in query 2 you are selecting all records ALONG WITH max(prev) that's meeting the conditions in the where and group by clause.
Basically, query 1 and query 2 have completely different where clauses. Hope this explains the missing records from query 1.

Answer 5

Your query #1 will certainly fail to return a row for a given pobj_name where maximum prev for that name does not correspond to a revision currently in the database. That could perhaps happen if a revision was skipped or if its row was deleted.
Your Query #2 does not suffer Query #1's limitation, and it may perform better on account of avoiding a correlated subquery. It would be inappropriate, however, if you wanted more data than just pobj_name and aggregate functions of the groups. And by the way, there's no point in including prev in the ORDER BY clause, since pobj_name will already be unique to each result row.
Overall, if the two queries happen to return similar results then that is a matter of the details of the data, not of the queries. They arrive at their results completely differently.