Github user mgaido91 commented on the issue:
https://github.com/apache/spark/pull/19811
@kiszk yes, I know, but I meant a different thing. I will try to explain
with an example, let me know if I am not clear enough.
Since we are not defining everything as an array and we are initializing
everything which doesn't go in an array before everything which goes into an
array, currently, the following situation can happen:
- we define a variable (`v1`) which goes into an array;
- then we define a variable (`v2DependsOnV1`) which doesn't go into any
array, but which relies on the previous variable;
Before this PR, we were ensuring that the order of the initialization was
the same of the declarations, then it would have been:
```
// here we init v1
// here we init v2DependsOnV1
```
After the PR, in the above condition, we don't maintain the same order and
we would have:
```
// here we init v2DependsOnV1
// here we init v1
```
which is a wrong situation.
Maybe this is a condition which never happens, ie. we don't have variables
which depends on others during the initialization, thus maybe I am worried for
a situation which never happens, but this is my question: are we sure this is
not a problem?
Thanks.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
