Shiakaron wrote:
Ok so, here is my example to prove that `FilesBeforeFolders` actually gives a
unique ordering which cannot be achieved with `IncludeCategories`. Consider
these two includes:
```
#include "net/http/client.h"
#include "net/socket.h"
```
`FilesBeforeFolders=true` would order them such that `"net/socket.h"` is before
`"net/http/client.h"` (great).
Now consider the provided IncludeCategories (ignoring the relative path bits):
```
- Regex: '^".*/'
Priority: 10
SortPriority: 11
CaseSensitive: false
- Regex: '^".*'
Priority: 10
SortPriority: 10
CaseSensitive: false
```
Because 'h' < 's' alphabetically, `"net/http/client.h"` would be before
`"net/socket.h"` (not what we want). This is because `^".*/` is a flat
alphabetical sort over a tree.
Now, as far as I understand, you are suggesting to add more patterns to deal
with the extra depth in the project, which will indeed solve it for this
minimal example - and this is what I have done in the config I have provided.
However, doing that breaks the inter-directory grouping the moment two
top-level directories both contain files and subdirectories.
Now consider the following includes:
```
#include "net/http/client.h"
#include "net/socket.h"
#include "io/file/reader.h"
#include "io/stream.h"
```
With `FilesBeforeFolders=true` we would get the following order:
```
#include "io/stream.h"
#include "io/file/reader.h"
#include "net/socket.h"
#include "net/http/client.h"
```
whereas with our new `IncludeCategories` we would get the following order:
```
#include "io/stream.h"
#include "net/socket.h"
#include "io/file/reader.h"
#include "net/http/client.h"
```
To achieve the same ordering we would instead need to enumerate every directory
by name which, I believe, is not a reasonable config for anyone to be expected
to have. I hope this is enough to convince you @HazardyKnusperkeks
Thank you for the time you have spent on this so far.
https://github.com/llvm/llvm-project/pull/208954
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
