MySQL CASE WHEN is not working as expected with user-defined variables?

i need help please with the following query. I am pulling zip codes (plz) along with user id (uid) from DB and ORDER them by zip (plz) and uid. Also I use a user-defined variable @prev to check if previous plz is different from current plz. In this CASE column rownum should decrease starting user-defined variable @row by 1.

Here is my query for this ..

SELECT   
  @prev as previous, @prev := u.plz as current, 
  CASE 
    WHEN @prev is null THEN @row := @row -1 
    WHEN @prev <> u.plz THEN @row := @row -1
    ELSE @row END AS rownum, 
  uid

FROM             
  (SELECT @prev := NULL, @row := 1001) as r, user as u

LEFT JOIN 
  geodb_locations ON u.plz = geodb_locations.id 
WHERE 
  u.plz IN(29386,30013,29271,30406) AND
ORDER BY 
  FIELD(u.plz,29386,30013,29271,30406), 
  uid DESC

Output

+----------+---------+--------+-------+
| previous | current | rownum | uid   |
+----------+---------+--------+-------+
| NULL     |   29386 |   1001 | 92014 | <-- @prev <> u.plz why NOT decreasing `rownum`
| 29386    |   29386 |   1001 | 87223 |
| 29386    |   29386 |   1001 | 83160 |
| 29386    |   29386 |   1001 | 81322 |
| 29386    |   29386 |   1001 | 81293 |
| 29386    |   29386 |   1001 |  3152 |
| 29386    |   29386 |   1001 |  1939 |
| 29386    |   30013 |   1001 | 65193 | <-- @prev <> u.plz why NOT decreasing `rownum`
| 30013    |   29384 |   1001 | 12883 |
| 29384    |   29385 |   1001 | 32667 |
| 29385    |   29232 |   1001 | 86641 |
| 29232    |   29481 |   1001 | 94521 |
| 29481    |   29481 |   1001 | 33950 |
| 29481    |   29481 |   1001 | 33926 |
| 29481    |   29667 |   1001 | 16027 |
| 29667    |   29667 |   1001 |   496 |
| 29667    |   29355 |   1001 | 96013 |
+----------+---------+--------+-------+

Expected output

+----------+---------+--------+-------+
| previous | current | rownum | uid   |
+----------+---------+--------+-------+
| NULL     |   29386 |   1000 | 92014 | <-- decrease @row `rownum` by one
| 29386    |   29386 |   1000 | 87223 |
| 29386    |   29386 |   1000 | 83160 |
| 29386    |   29386 |   1000 | 81322 |
| 29386    |   29386 |   1000 | 81293 |
| 29386    |   29386 |   1000 |  3152 |
| 29386    |   29386 |   1000 |  1939 |
| 29386    |   30013 |    999 | 65193 | <-- decrease @row `rownum` by one 
| 30013    |   29384 |    998 | 12883 |
| 29384    |   29385 |    997 | 32667 |
| 29385    |   29232 |    996 | 86641 |
| 29232    |   29481 |    995 | 94521 |
| 29481    |   29481 |    995 | 33950 |
| 29481    |   29481 |    995 | 33926 |
| 29481    |   29667 |    994 | 16027 |
| 29667    |   29667 |    994 |   496 |
| 29667    |   29355 |    993 | 96013 |
+----------+---------+--------+-------+

Why is CASE WHEN not working like expected? Where is my mistake in thinking?
Thanks in advance ..

The order of output expressions evaluation is not documented or specified. In general the evaluating order matches the text.

Look. You assign @prev := u.plz and then try to compare these values – but they are already equal due to assigning.

SELECT   
  @prev as previous,                          -- @prev stores "old" value
  @prev := u.plz as current,                  -- @prev is re-assigned to u.plz
  CASE 
    WHEN @prev is null THEN @row := @row -1   -- @prev is equal to u.plz, FALSE
    WHEN @prev <> u.plz THEN @row := @row -1  -- @prev is equal to u.plz, FALSE
    ELSE @row END AS rownum,                  -- this alternative is evaluated always
  uid

Place this assign AFTER the CASE expression – in this case CASE will use old (taken from previous row calculations) value.

SELECT   
  @prev as previous,                          -- @prev stores "old" value
  CASE 
    WHEN @prev is null THEN @row := @row -1   -- "old" value is used for conditions
    WHEN @prev <> u.plz THEN @row := @row -1 
    ELSE @row END AS rownum,                 
  @prev := u.plz as current,                  -- and now @prev is re-assigned to u.plz
  uid