[PATCH] perf hist: Fix width calculation in hpp__fmt()

[PATCH] perf hist: Fix width calculation in hpp__fmt()

[Dmitry Vyukov]

hpp__width_fn() round up width to length of the field name,
hpp__fmt() should do it too. Otherwise, the numbers may
end up unaligned if the field name is long.

Signed-off-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ian Rogers <irogers@google.com>
Cc: linux-perf-users@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
---
 tools/perf/ui/hist.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c
index e5491995adf08..34fda1d5eccb4 100644
--- a/tools/perf/ui/hist.c
+++ b/tools/perf/ui/hist.c
@@ -121,7 +121,7 @@ int hpp__fmt(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 	     const char *fmtstr, hpp_snprint_fn print_fn,
 	     enum perf_hpp_fmt_type fmtype)
 {
-	int len = fmt->user_len ?: fmt->len;
+	int len = max(fmt->user_len ?: fmt->len, (int)strlen(fmt->name));
 
 	if (symbol_conf.field_sep) {
 		return __hpp__fmt(hpp, he, get_field, fmtstr, 1,

base-commit: 09a0fa92e5b45e99cf435b2fbf5ebcf889cf8780
-- 
2.47.1.613.gc27f4b7a9f-goog

Re: [PATCH] perf hist: Fix width calculation in hpp__fmt()

[James Clark]

On 08/01/2025 6:59 am, Dmitry Vyukov wrote:
> hpp__width_fn() round up width to length of the field name,
> hpp__fmt() should do it too. Otherwise, the numbers may
> end up unaligned if the field name is long.
> 
> Signed-off-by: Dmitry Vyukov <dvyukov@google.com>
> Cc: Namhyung Kim <namhyung@kernel.org>
> Cc: Ian Rogers <irogers@google.com>
> Cc: linux-perf-users@vger.kernel.org
> Cc: linux-kernel@vger.kernel.org
> ---
>   tools/perf/ui/hist.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c
> index e5491995adf08..34fda1d5eccb4 100644
> --- a/tools/perf/ui/hist.c
> +++ b/tools/perf/ui/hist.c
> @@ -121,7 +121,7 @@ int hpp__fmt(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
>   	     const char *fmtstr, hpp_snprint_fn print_fn,
>   	     enum perf_hpp_fmt_type fmtype)
>   {
> -	int len = fmt->user_len ?: fmt->len;
> +	int len = max(fmt->user_len ?: fmt->len, (int)strlen(fmt->name));
>   
>   	if (symbol_conf.field_sep) {
>   		return __hpp__fmt(hpp, he, get_field, fmtstr, 1,
> 
> base-commit: 09a0fa92e5b45e99cf435b2fbf5ebcf889cf8780

This seems to right align the field correctly with a long name. Although 
it's worth noting I had to add a long name to make it obvious.

Reviewed-by: James Clark <james.clark@linaro.org>

I saw some other existing issue with small column-width arguments which 
is changed slightly by this patch. But I don't think it's worse:


$ perf report --fields 
overhead,overhead_sys,overhead_us,overhead_guest_us,dso 
--column-widths=3,3,3,3,3

After:
------
Overhead       sys       usr  guest usr long name  Shared Object
   63.78%     0.00%    63.78%                0.00%  ld-linux-x86-64.so

After, width=3:
--------------
Overhead       sys       usr  guest usr long name  Sha
   63.78%   0.00%     63.78%                 0.00%  ld-

Before:
---------------
Overhead       sys       usr  guest usr long name  Shared Object
   63.78%     0.00%    63.78%      0.00%            ld-linux-x86-64.so

Before, width=3:
---------------
Overhead       sys       usr  guest usr long name  Sha
  63.78%    0.00%     63.78%    0.00%               ld-

Re: [PATCH] perf hist: Fix width calculation in hpp__fmt()

[James Clark]

On 08/01/2025 5:26 pm, James Clark wrote:
> 
> 
> On 08/01/2025 6:59 am, Dmitry Vyukov wrote:
>> hpp__width_fn() round up width to length of the field name,
>> hpp__fmt() should do it too. Otherwise, the numbers may
>> end up unaligned if the field name is long.
>>
>> Signed-off-by: Dmitry Vyukov <dvyukov@google.com>
>> Cc: Namhyung Kim <namhyung@kernel.org>
>> Cc: Ian Rogers <irogers@google.com>
>> Cc: linux-perf-users@vger.kernel.org
>> Cc: linux-kernel@vger.kernel.org
>> ---
>>   tools/perf/ui/hist.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c
>> index e5491995adf08..34fda1d5eccb4 100644
>> --- a/tools/perf/ui/hist.c
>> +++ b/tools/perf/ui/hist.c
>> @@ -121,7 +121,7 @@ int hpp__fmt(struct perf_hpp_fmt *fmt, struct 
>> perf_hpp *hpp,
>>            const char *fmtstr, hpp_snprint_fn print_fn,
>>            enum perf_hpp_fmt_type fmtype)
>>   {
>> -    int len = fmt->user_len ?: fmt->len;
>> +    int len = max(fmt->user_len ?: fmt->len, (int)strlen(fmt->name));
>>       if (symbol_conf.field_sep) {
>>           return __hpp__fmt(hpp, he, get_field, fmtstr, 1,
>>
>> base-commit: 09a0fa92e5b45e99cf435b2fbf5ebcf889cf8780
> 
> This seems to right align the field correctly with a long name. Although 
> it's worth noting I had to add a long name to make it obvious.
> 
> Reviewed-by: James Clark <james.clark@linaro.org>
> 
> I saw some other existing issue with small column-width arguments which 
> is changed slightly by this patch. But I don't think it's worse:
> 

Ah maybe I was looking at it wrong, before and after are the same with 
width=3, apart from the intended fix. False alarm.

> 
> $ perf report --fields 
> overhead,overhead_sys,overhead_us,overhead_guest_us,dso --column- 
> widths=3,3,3,3,3
> 
> After:
> ------
> Overhead       sys       usr  guest usr long name  Shared Object
>    63.78%     0.00%    63.78%                0.00%  ld-linux-x86-64.so
> 
> After, width=3:
> --------------
> Overhead       sys       usr  guest usr long name  Sha
>    63.78%   0.00%     63.78%                 0.00%  ld-
> 
> Before:
> ---------------
> Overhead       sys       usr  guest usr long name  Shared Object
>    63.78%     0.00%    63.78%      0.00%            ld-linux-x86-64.so
> 
> Before, width=3:
> ---------------
> Overhead       sys       usr  guest usr long name  Sha
>   63.78%    0.00%     63.78%    0.00%               ld-

Re: [PATCH] perf hist: Fix width calculation in hpp__fmt()

[Namhyung Kim]

On Wed, 08 Jan 2025 07:59:34 +0100, Dmitry Vyukov wrote:

> hpp__width_fn() round up width to length of the field name,
> hpp__fmt() should do it too. Otherwise, the numbers may
> end up unaligned if the field name is long.
> 
> 

Applied to perf-tools-next, thanks!

Best regards,
Namhyung