Changing default setting of bash's hash table?

[Roman Rakus]

Maybe the subject is a bit misleading, I will clarify it.

Bash is using hash table to remember locations of executed commands. 
Whenever you try to run a command bash looks in hash table. When the 
command is found in table then bash will you full path name as it is in 
the table.

However there is a problem when the command moved (or is deleted). Bash 
by default is not checking if the command is really on the location. But 
there is bash option that will force bash to check if the command really 
exists. Man page says:
checkhash
If set, bash checks that a command found in the hash ta‐
ble exists before trying to execute it. If a hashed
command no longer exists, a normal path search is per‐
formed.

I have a question, if it is worth to enable this option by default? It 
will not confuse some people, but can increase disk searching. Comments 
welcome.

RR

[Roberto Ragusa]

On 08/19/2011 01:24 PM, Roman Rakus wrote:
>  I have a question, if it is worth to enable this option by default?
> It 
>  will not confuse some people, but can increase disk searching. Comments 
>  welcome. 
Confusion can only happen in this particular case, I think:
1) you run a command and the cache is populated
2) you delete the command
3) you try to run the command again, in the same shell
4) you expect bash to find the command in another path, but get an error instead
Correct me if I'm wrong.

I remember having had to use "hash -r" one or two times in my life
for this kind of issues.

>  RR 
Hey, I'm RR too. :-)

-- 
   Roberto Ragusa    mail at robertoragusa.it

[Maciej Małecki]

2011/8/19 Roberto Ragusa <mail(a)robertoragusa.it&gt;
>  On 08/19/2011 01:24 PM, Roman Rakus wrote:
>  > I have a question, if it is worth to enable this option by default? It
>  > will not confuse some people, but can increase disk searching. Comments
>  > welcome.
>
>  Confusion can only happen in this particular case, I think:
>  1) you run a command and the cache is populated
>  2) you delete the command
>  3) you try to run the command again, in the same shell
>  4) you expect bash to find the command in another path, but get an error instead
>  Correct me if I'm wrong.
>
>  I remember having had to use "hash -r" one or two times in my life
>  for this kind of issues.
>  
Not worth it, one can always use which to verify if command is gone or
is bash is going mad.
(It happened to me one, I used yum remove with -y and didn't expect
one dependency).

>  > RR
>  Hey, I'm RR too. :-) 
Well, does MM count?

--
Greetings,
Maciej Małecki

[Ville Skyttä]

On 08/20/2011 12:09 AM, Maciej Małecki wrote:

>  Not worth it, one can always use which to verify if command is gone
> or
>  is bash is going mad. 
+1

[Till Maas]

On Fri, Aug 19, 2011 at 01:24:37PM +0200, Roman Rakus wrote:

>  I have a question, if it is worth to enable this option by default?
> It 
>  will not confuse some people, but can increase disk searching. Comments 
>  welcome. 
How can it increase disk searching in case the program is still there?
It needs to be fetched from the disk anyhow when it is run. Or am I
missing something here?

Regards
Till

[Maciej Małecki]

2011/8/20 Till Maas <opensource(a)till.name&gt;:
>  How can it increase disk searching in case the program is still
> there?
>  It needs to be fetched from the disk anyhow when it is run. Or am I
>  missing something here? 
It's just one stat with every command execution.
If you're curious how it looks internally, findcmd.c:71 and findcmd.c:327.
Function which gets executed is file_status, defined in findcmd.c:84.
--
Greetings,
Maciej Małecki

[Kevin Kofler]

Roman Rakus wrote:
>  Maybe the subject is a bit misleading, I will clarify it.
>  
>  Bash is using hash table to remember locations of executed commands.
>  Whenever you try to run a command bash looks in hash table. When the
>  command is found in table then bash will you full path name as it is in
>  the table.
>  
>  However there is a problem when the command moved (or is deleted). Bash
>  by default is not checking if the command is really on the location. But
>  there is bash option that will force bash to check if the command really
>  exists. Man page says:
>  checkhash
>  If set, bash checks that a command found in the hash ta‐
>  ble exists before trying to execute it. If a hashed
>  command no longer exists, a normal path search is per‐
>  formed.
>  
>  I have a question, if it is worth to enable this option by default? It
>  will not confuse some people, but can increase disk searching. Comments
>  welcome. 
It will only help if the first match in the search path is removed, it will 
still not do the right thing if a new match is prepended to the search path.

IMHO, this whole hashing should not be done in interactive shells at all. 
The bottleneck is going to be the user running the commands anyway. I can 
see how it speeds up scripts, but it's just confusing and useless in 
interactive operation.

        Kevin Kofler