On 10/23/18 11:46 AM, francky.leyn@telenet.be wrote:
> Hello, 
> 
> I just destroyed 3 days of scaninng work. This was the command: 
> 
> ls p.*.png | sort -t. -k2 -n | gawk '{ printf("mv '\''%s'\'' '\''p. %d.png'\''\n", $0, NR+6) }' | bash [ mailto:bug-coreutils@gnu.org ] 
> 
> The problem was that the sequence of mv command resulted in one .png file: 
> 
> mv 'p. 123.png' 'p. 124.png' 
> mv 'p. 124.png' 'p. 125.png' 
> mv 'p. 126.png' 'p. 127.png' 
> 
> The result is that you end up with one file: 'p. 127.png'. 

outch, sorry for your loss.

> When you look at the commands (without the trailing bash), you think it's ok because you think declaratively. However, this thing works imperative/procedural. 

As you said, the problem originated from a logic error in the counting
for the new names of the files.
Didn't you run that without the executing "| bash" first?

I did, but my coin hasn't fallen. What I should have done, is after seeing the result when executing without the "| bash", is change the command to:

ls p.*.png | sort -t. -k2 -n | gawk '{ printf("mv -i '\''%s'\'' '\''p. %d.png'\''\n", $0, NR+6) }' | tac | bash

> I would not have lost all my files if I would have used "mv -i": 
> 
> ls p.*.png | sort -t. -k2 -n | gawk '{ printf("mv -i '\''%s'\'' '\''p. %d.png'\''\n", $0, NR+6) }' | bash 

No, because once you would have confirmed the -i prompt, the same would have happened.
The point is to recognize that file 123 is first renamed to 124, then 124 is renamed
to 125, and so on.

> ============================== 
> 
> With the rm command it is the same. It deletes/removes, and the file is gone forever. No tracing back option like a recycle bin under Windows. 
> 
> In this mail I make a plei for a more protective manner for the coreutils. The "-i" options should dissapear altoghether and being the default. Instead there should be options to force execution without confirmation. 

The GNU coreutils are a low-level tool set which adheres to certain
standards and provides certain compatibility with other implementations.
We cannot (and therefore will not) simply change such behavior.
Furthermore, -i is not effective in scripts without stdin (in this case
stdin came from the pipe).
Finally, as written above, -i would *not* have helped you unless you
would have seen the logical mistake.

As such, I'm afraid we can not do something at this point.

> Also: 
> 
> 
>     * There should be a recycle bin for rm. 
>     * There should be a call back facility to undo the latest action. In Windowds this is ctrl z. 

Different topic.
I think this has been already discussed.  Again, the GNU coreutils as command line
tools are quite low-level, and things like a recycle bin belongs to desktop apps.

I once heard about a "trish cli" or something package that implements the rm in the Windows way. I was command line driven. But I was unable to install the package. Do you perhaps know the exact name?

I'd compare it with using a sharp Japanese kitchen knife: of course one can cut oneself
into the fingers, but if one uses it as it's designed for, then it can do things
one can't with a blunt, old steel knife (although the latter may or may not have
its merits regarding other aspects).

Do you know a solution to implement the Windows ctrl z feature?

Have a nice day,
Berny

Beantwoorden - Iedereen antwoorden - Doorsturen - Meer a