Re: Interest in selectively deleting history entries.

[Daniel J Sebald]

On 09/21/2014 12:50 PM, Philip Nienhuis wrote:
> Daniel Sebald wrote
> > I'm curious what people's thoughts are on the ability to selectively
> > delete individual lines of history.  What I see it used for, in
> > particular, is removing a password the user accidentally types and hits
> > return.  That could happen if the user thinks the account is locked but
> > it is actually the monitor's screen saver.
>
> Yep - recognized, but maybe not so urgent as -I suppose- it rarely happens.

It's not common, but at the same time I wouldn't call it rare.  When 
someone (I) concentrate on a problem, typing away, it's pretty easy to 
type a screen password because it's something done so often, almost 
habitual.


> And it doesn't only hold for Octave but also for shells etc.

Yes, I know.  At least in that case one can clear the screen fairly 
quickly.  With Octave GUI one's password would sit plain as day in the 
history window.  And as you pointed out regarding shells, with so many 
windows open the history is fragmented and only good for about the last 
five lines.  But with Octave, people might have a complex sequence of 
commands they want to remember/retrieve down the road.


> Octave's
> history file is an editable text file so there is a way out.

Yes, but hardly convenient to save, exit, edit, launch.


> While we are at history, another nice thing would be the ability to delete
> multiple identical history entries and only retain the last one - ideally as
> a configurable setting. At times I have countless "dbstep"/"dbnext" entries
> that just clobber up the history.

I hear you there.  Keep hitting up arrow and re-running a command and 
there's a big string of the same command in history.  I see history as a 
very configurable thing.  There is the way that lines are stored in 
history.  There is also the way they are displayed and retrieved.  For 
example, one could save every line in history while only display (and 
step through) lines that are contiguously unique.  Say I hit the 
up-arrow and retrieve the previous command three times.  History might 
be displayed as:

> x = 1;
> y = x + 1;
> y = y^2;
> y = y^2;
> y = y^2;
> history

...
1003 x = 1;
1004 y = x + 1;
1007 y = y^2;

That is, every line has a unique number, but not all are displayed.


> If the relevant code is in .m-shape (rather than C++) I could have a look
> myself.

history is a builtin command.  It would need to be C++.


> > I think I saw that the internal list code has the ability to delete
> > individual lines.  So that part is straightforward.  The question is
> > what to do with the numbering.  If a line is deleted, should there just
> > be a hole in the numbering sequence where the line used to be?  I think
> > it is probably preferred to not change the number of a line in history
> > once it is assigned.  So that would make an extra level of mapping the
> > list number to a command-line number.
> >
> > Is there a use for numbered history if there is nothing in the history
> > command that can act on the line number?
>
> I suppose the history "memory" is limited?
> How does the numbering work? Does it start counting from the very first item
> in the history list (after having read<history_file>) or just at the first
> command of the newly started Octave session?

The history list loads at 1, so when starting the session it's likely 
your first command will be number 1001, if your history length is 1000.


> In case of the former, what happens when two or more Octave sessions run
> simultaneously on the same box (happens regularly here) and write their
> history to file (same file!) when closed down?
>
> All of this makes me assume that the history entry number makes most sense
> as a per-session counter. But that's just me.

It is per-session, yes.


> Thinking about it, relying on a history entries' number, while without doubt
> potentially useful, sounds quite fragile to me.

I agree, but if one were to have a line-deleting feature, the history 
would have to have a number associated with it...and not change, 
otherwise that gets confusing.  It would be nice to have absolute 
history line referencing

"history -D 1003"

and relative number referencing, say,

"history -D -2"

because if one is working in the GUI it's easy to see the history window 
and count backward from the present command.

And I've always been philosophically conflicted with whether the history 
command itself should go into history.  If I delete a line of code, do I 
really want "history -D -2" to take it's place?  Just seems defeating.

History with the pager isn't as convenient as it could be.  Is there a 
way to make the pager go to the bottom (most recent) of the list rather 
than start at the top (oldest) when running the history command?

Sounds picky, but the things that are used most often should function 
the most efficiently.

Dan



> Philip
>
>
>
>
> --
> View this message in context: 
> http://octave.1599824.n4.nabble.com/Interest-in-selectively-deleting-history-entries-tp4666612p4666615.html
> Sent from the Octave - Maintainers mailing list archive at Nabble.com.

--

Dan Sebald
email: daniel(DOT)sebald(AT)ieee(DOT)org
URL: http://www(DOT)dansebald(DOT)com