Обсуждение: Re: [PATCH] analyze: move elevel calculation into do_analyze_rel()

Hi hackers,

I reworked the patch slightly.

Since do_analyze_rel() is static, removing the extra elevel argument is
local to analyze.c only. No behavioral change is intended.

--

Best regards,

Lev Nikolaev,

Tantor Labs LLC,

https://tantorlabs.com/

Hi hackers,

While reading analyze.c in current master, I noticed a small cleanup opportunity in the ANALYZE code path.

Currently, analyze_rel() computes elevel from params.options, but does not use it directly. It only passes the value down to do_analyze_rel(). At the same time, do_analyze_rel() already receives params and derives verbose from params.options for its own purposes.

The attached patch moves the elevel calculation into do_analyze_rel() and removes the extra elevel argument from its declaration, definition, and call sites.

No behavioral change is intended here. The logging level selection remains the same; the calculation is just moved closer to the actual use sites.

Comments and feedback would be appreciated.

 

--

Best regards,

Lev Nikolaev,

Tantor Labs LLC,

https://tantorlabs.com/

On 3/31/26 7:56 PM, Lev Nikolaev wrote:
> I reworked the patch slightly.

This small cleanup makes sense to me since do_analyze_rel() already 
looks at if VACOPT_VERBOSE is set meaning related code is grouped closer 
after this small refactoring.

The patch no longer applied so I rebased it.

Andreas

On Thu, Apr 2, 2026 at 8:29 AM Andreas Karlsson <andreas@proxel.se> wrote:
> On 3/31/26 7:56 PM, Lev Nikolaev wrote:
> > I reworked the patch slightly.
>
> This small cleanup makes sense to me since do_analyze_rel() already
> looks at if VACOPT_VERBOSE is set meaning related code is grouped closer
> after this small refactoring.
>
> The patch no longer applied so I rebased it.

Sorry, but -1 from me because this change would lead to doing the same
setup repeatedly when analyzing inheritance trees, which is not great.

Best regards,
Etsuro Fujita