_pyrepl/pager.py: call less with --clear-screen

[StanFromIreland]

Since this is only a cosmetic change, I don't think it qualifies for a news entry.

Currently, text is written bottom-to-top, resulting in odd placement for copyright():

This patch fixes that:

By passing --clear-screen (-c) to less, which "causes full screen repaints to be painted from the top line down."

[pablogsal]

Thank you for your contribution! :)

[StanFromIreland]

Thanks for your merge!

[skirpichev]

Sorry, @StanFromIreland, it doesn't look as a cosmetic change. IMO, it's a regression, which introduces rather unhelpful behavior of the REPL.

Here old behavior:

So, for small docstring - you can see all nearby repl interaction without scrolling.

Now it's not the case:

BTW, in the real world - my console size would be much bigger. So, more wasted space...

Maybe we could solve your issue (if it exists at all) in a different way? BTW, why to keep pager at all for small texts? I think that git tools do this per default only if text doesn't fit to the screen (-F option).

The patch:

diff --git a/Lib/_pyrepl/pager.py b/Lib/_pyrepl/pager.py
index 92afaa5933a..f937d7973d8 100644
--- a/Lib/_pyrepl/pager.py
+++ b/Lib/_pyrepl/pager.py
@@ -138,7 +138,7 @@ def pipe_pager(text: str, cmd: str, title: str = '') -> None:
         '.'
         '?e (END):?pB %pB\\%..'
         ' (press h for help or q to quit)')
-    env['LESS'] = '-RcmPm{0}$PM{0}$'.format(prompt_string)
+    env['LESS'] = '-F -RmPm{0}$PM{0}$'.format(prompt_string)
     proc = subprocess.Popen(cmd, shell=True, stdin=subprocess.PIPE,
                             errors='backslashreplace', env=env)
     assert proc.stdin is not None

The behavior for git tools could be customized, i.e. they respect the LESS variable. Maybe we also should introduce some opportunity to opt-out from defaults?

[StanFromIreland]

I'm afraid I can't reproduce your shown old behaviour, I always get the pager anyway, testing on 3.14 (which didn't get this fix):

Maybe this is down to different lesses? I'm not sure what the output on Mac would be.

[skirpichev]

I'm afraid I can't reproduce your shown old behaviour, I always get the pager anyway

You are right, this happens only if you near of the bottom of the screen. Though, it's a sort of common case in a long REPL session.

BTW, with added -F option - this happens always. Are there reasons to keep pager for a small texts? Terminals e.g. usually have own capabilities for text search and so on.

I'm not sure what the output on Mac would be.

No, it was linux.

[serhiy-storchaka]

This look like a regression to me. Previously, the text of the help remained on the screen after quiting the pager. Now it disappears.

I think this change should be reverted.

[serhiy-storchaka]

Hmm, it disappeared also in old Python versions. This was inconvenient sometimes.

[StanFromIreland]

Previously, the text of the help remained on the screen after quiting the pager.

I can't reproduce this in 3.14 either? Are you sure that wasn't a different change? Or again just small differences in less behaviour on different platforms 🙃.

[serhiy-storchaka]

I was wrong. Sorry.

[StanFromIreland]

Oh! I didn't see #146382 (comment) which must have been posted as I was writing, sorry!

[skirpichev]

Hmm, it disappeared also in old Python versions. This was inconvenient sometimes.

This was explained in the comment above: #146382 (comment)

Usually the prompt will be at the bottom of the screen - and then the pager will appear at the bottom, but the screen will not be cleared.

For this reason, I think that backporting #148322 not so important.

[serhiy-storchaka]

Anyway, I think this change should be reverted. See #148321 (comment).