Android SQLite 获得相对于其他列的最大值

Question

I have a table called 'tracked_exercises' which contains columns shown in the attached picture (exercise_name, band, reps, kgs)

I'd like to create a query which given an exercise_name returns all rows which have any of these conditions for that exercise_name

• Max reps with each unique band
• Max kgs at each reps with matching band(But not if max kgs are lower or equal than max kgs with higher reps. So if you have 2 reps with 6kg, and 3 reps with 6kg I do not want the 2 reps with 6kg back)

So in the example image if I give 'Muscle Up' I would get back

   | band   | reps | kgs |
   | red    | 10   | 0   |
   | purple | 15   | 0   |
   | No     | 5    | 0   |
   | No     | 4    | 10  |
   | No     | 2    | 15  |

I have been able to get the maximum reps for each unique band with the below query, but haven't managed to work out how to get rows with max kgs as well

@Query("SELECT * FROM tracked_exercises INNER JOIN (SELECT band, MAX(reps) AS Maxreps FROM tracked_exercises WHERE exercise_name =:name GROUP BY band ) topset ON tracked_exercises.band = topset.band AND tracked_exercises.reps = topset.Maxreps")
List<TrackedExercise> getPersonalRecords(String name);

Sorry if this is not clear, comment if you need more info

Example tracked_exercises table

Another example:

| band  | reps | kgs |
| No    | 3    | 2.5 |
| No    | 3    | 5   |
| No    | 3    | 6.25|
| No    | 10   | 0   |
| No    | 9    | 0   |
| No    | 4    | 5   |
| No    | 4    | 3.75|
| red   | 11   | 0   |
| red   | 10   | 0   |
| No    | 2    | 6.25|

I'd like the result:

| band  | reps | kgs |
| No    | 10   | 0   |
| red   | 11   | 0   |
| No    | 3    | 6.25|
| No    | 4    | 5   |

Edit: updated question to make it more clear

Answer 1

Use a CTE to get all the maximums for reps and kgs and then NOT EXISTS on the results:

with cte as (
  select t.band, t.reps, t.kgs
  from ( 
    select *,
      row_number() over (partition by band order by reps desc) rnreps,
      row_number() over (partition by band, reps order by kgs desc) rnkgs
    from tracked_exercises
  ) t  
  where t.rnreps = 1 or t.rnkgs = 1
)  
select c.* from cte c
where not exists (
  select 1 from cte
  where band = c.band and reps > c.reps and kgs >= c.kgs
)

See the demo .
Results:

| band | reps | kgs  |
| ---- | ---- | ---- |
| No   | 10   | 0    |
| No   | 4    | 5    |
| No   | 3    | 6.25 |
| red  | 11   | 0    |