[RFD/BUG?] git show with tree/commit

[RFD/BUG?] git show with tree/commit

[Michael J Gruber]

With current next, using any commit sha1:

$ git cat-file -p cab2cdadfda8e8e8631026443b11d3ed6e7ba517
tree 04050d37b1676ba8da277be1902513049b45413a
parent b673b5e7d121021f77768c736cd9f98d7b3b3967
parent 7b0d47b3b6b5b64e02a5aa06b0452cadcdb18355
...

$ git show cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
tree cab2cdadfda8e8e8631026443b11d3ed6e7ba517:

.gitattributes
.gitignore
.mailmap
...

$ git show 04050d37b1676ba8da277be1902513049b45413a:
tree 04050d37b1676ba8da277be1902513049b45413a

.gitattributes
.gitignore
.mailmap
...

While it's clear to me what's going on, I'm wondering whether it's a
good idea that "git show" says "tree" in front of the unresolved
treeish, whether it's a tree, a commit or something else. I think it's
pretty confusing. Alternatives would be:

tree <resolved tree id> # here: 040...
treeish <treeish>	# here: "treeish cab2c..."
tree <treeish>^{tree} 	# here: "tree cab2c^{tree}"

Cheers
Michael

Re: [RFD/BUG?] git show with tree/commit

[Junio C Hamano]

Michael J Gruber <git@drmicha.warpmail.net> writes:

> $ git show cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
> tree cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>
> .gitattributes
> .gitignore
> .mailmap
> ...
>
> While it's clear to me what's going on, I'm wondering whether it's a
> good idea that "git show" says "tree" in front of the unresolved
> treeish, whether it's a tree, a commit or something else. I think it's
> pretty confusing.

There is no "unresolved treeish" on the line that begins with
"tree", but I suspect it wasn't very clear to you because of the way
you gave the command its input; notice the line in question ends
with a colon, which is unfortunate, but it turns out that it is your
fault ;-).  Read on.

> Alternatives would be:
>
> tree <resolved tree id> # here: 040...
> treeish <treeish>	# here: "treeish cab2c..."
> tree <treeish>^{tree} 	# here: "tree cab2c^{tree}"

So, the three choices are

 (1) resolve the tree object name to 40-hex and show it as
     "tree <object name in hex>"

 (2) given an object that is not a tree, show it as "treeish <object
     name>" 

 (3) do not do anything fancy, just show it as "tree <object name>"
     using what the user gave us.

I think the current output is doing the third one (notice the colon
at the end of the line).

    $ git show master: | head -n 1
    $ git show master^{tree} | head -n 1
    $ git show cab2cdadf: | head -n 1

Re: [RFD/BUG?] git show with tree/commit

[Michael J Gruber]

Junio C Hamano venit, vidit, dixit 12.04.2016 18:26:
> Michael J Gruber <git@drmicha.warpmail.net> writes:
> 
>> $ git show cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>> tree cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>>
>> .gitattributes
>> .gitignore
>> .mailmap
>> ...
>>
>> While it's clear to me what's going on, I'm wondering whether it's a
>> good idea that "git show" says "tree" in front of the unresolved
>> treeish, whether it's a tree, a commit or something else. I think it's
>> pretty confusing.
> 
> There is no "unresolved treeish" on the line that begins with
> "tree", but I suspect it wasn't very clear to you because of the way
> you gave the command its input; notice the line in question ends
> with a colon, which is unfortunate, but it turns out that it is your
> fault ;-).  Read on.

Guess what, I know the difference between "git show <commit>" and git
show <commit>:"...

>> Alternatives would be:
>>
>> tree <resolved tree id> # here: 040...
>> treeish <treeish>	# here: "treeish cab2c..."
>> tree <treeish>^{tree} 	# here: "tree cab2c^{tree}"
> 
> So, the three choices are
> 
>  (1) resolve the tree object name to 40-hex and show it as
>      "tree <object name in hex>"
> 
>  (2) given an object that is not a tree, show it as "treeish <object
>      name>" 
> 
>  (3) do not do anything fancy, just show it as "tree <object name>"
>      using what the user gave us.

(3) ist not quite what I'm after, but I guess when "git show" receives
"<commit>:" there is no way it could trace back that tree reference to
the commit. In fact, (3) is what we have now.

> I think the current output is doing the third one (notice the colon
> at the end of the line).
> 
>     $ git show master: | head -n 1
>     $ git show master^{tree} | head -n 1
>     $ git show cab2cdadf: | head -n 1
> 

I confess that I failed to copy the trailing ":" from the original
output... So that one is my fault.

In fact, while specifying "<commit>:" knowingly to show the tree, I
interpreted the ":" in the output as a simple colon introducing the
following output, at least at first glance.

I guess (1) would be the only option that would help (because
"<commit>:" is a tree, not a non-tree treeish).

Michael

Re: [RFD/BUG?] git show with tree/commit

[Michael J Gruber]

Michael J Gruber venit, vidit, dixit 13.04.2016 10:02:
> Junio C Hamano venit, vidit, dixit 12.04.2016 18:26:
>> Michael J Gruber <git@drmicha.warpmail.net> writes:
>>
>>> $ git show cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>>> tree cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>>>
>>> .gitattributes
>>> .gitignore
>>> .mailmap
>>> ...
>>>
>>> While it's clear to me what's going on, I'm wondering whether it's a
>>> good idea that "git show" says "tree" in front of the unresolved
>>> treeish, whether it's a tree, a commit or something else. I think it's
>>> pretty confusing.
>>
>> There is no "unresolved treeish" on the line that begins with
>> "tree", but I suspect it wasn't very clear to you because of the way
>> you gave the command its input; notice the line in question ends
>> with a colon, which is unfortunate, but it turns out that it is your
>> fault ;-).  Read on.
> 
> Guess what, I know the difference between "git show <commit>" and git
> show <commit>:"...
> 
>>> Alternatives would be:
>>>
>>> tree <resolved tree id> # here: 040...
>>> treeish <treeish>	# here: "treeish cab2c..."
>>> tree <treeish>^{tree} 	# here: "tree cab2c^{tree}"
>>
>> So, the three choices are
>>
>>  (1) resolve the tree object name to 40-hex and show it as
>>      "tree <object name in hex>"
>>
>>  (2) given an object that is not a tree, show it as "treeish <object
>>      name>" 
>>
>>  (3) do not do anything fancy, just show it as "tree <object name>"
>>      using what the user gave us.
> 
> (3) ist not quite what I'm after, but I guess when "git show" receives
> "<commit>:" there is no way it could trace back that tree reference to
> the commit. In fact, (3) is what we have now.
> 
>> I think the current output is doing the third one (notice the colon
>> at the end of the line).
>>
>>     $ git show master: | head -n 1
>>     $ git show master^{tree} | head -n 1
>>     $ git show cab2cdadf: | head -n 1
>>
> 
> I confess that I failed to copy the trailing ":" from the original
> output... So that one is my fault.
> 
> In fact, while specifying "<commit>:" knowingly to show the tree, I
> interpreted the ":" in the output as a simple colon introducing the
> following output, at least at first glance.
> 
> I guess (1) would be the only option that would help (because
> "<commit>:" is a tree, not a non-tree treeish).
> 
> Michael

... adding to that, "git show foo" shows

"commit <sha1 of foo>" for a commit object
"tree <verbatim foo>" for a tree object

nothing like that for tag objects and blobs (just the object contents)

So for 2 out of the 4 object types, it shows a "header line", and both
of them show the specified argument differently (resolved to sha1 resp.
verbatim). It's that inconsistency that I find confusing.

Michael

Re: [RFD/BUG?] git show with tree/commit

[Jeff King]

On Tue, Apr 12, 2016 at 05:33:33PM +0200, Michael J Gruber wrote:

> $ git show cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
> tree cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
> 
> .gitattributes
> .gitignore
> .mailmap
> ...

As Junio pointed out, the colon here is not syntactic, but from the
original object name you gave it. But here...

> $ git show 04050d37b1676ba8da277be1902513049b45413a:
> tree 04050d37b1676ba8da277be1902513049b45413a
> 
> .gitattributes
> .gitignore
> .mailmap
> ...

...it is missing. Did you mean to omit it from the command-line?

-Peff

Re: [RFD/BUG?] git show with tree/commit

[Junio C Hamano]

Jeff King <peff@peff.net> writes:

> On Tue, Apr 12, 2016 at 05:33:33PM +0200, Michael J Gruber wrote:
>
>> $ git show cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>> tree cab2cdadfda8e8e8631026443b11d3ed6e7ba517:
>> 
>> .gitattributes
>> .gitignore
>> .mailmap
>> ...
>
> As Junio pointed out, the colon here is not syntactic, but from the
> original object name you gave it. But here...
>
>> $ git show 04050d37b1676ba8da277be1902513049b45413a:
>> tree 04050d37b1676ba8da277be1902513049b45413a
>> 
>> .gitattributes
>> .gitignore
>> .mailmap
>> ...
>
> ...it is missing. Did you mean to omit it from the command-line?

I suspect that is cut-and-paste something.  I see:

    $ git show 04050d37b1676ba8da277be1902513049b45413a: | head -n1
    tree 04050d37b1676ba8da277be1902513049b45413a:
    $ git show 04050d37b1676ba8da277be1902513049b45413a | head -n1
    tree 04050d37b1676ba8da277be1902513049b45413a

which is expected.

It indeed becomes confusing when you give "<treeish>:" from the
command line, as it looks as if there is conceptually a list of tree
contents that is shown like so:

    tree tree1:
    contents of tree1

    tree tree2:
    contents of tree2
    ...

and we are showing only the first one.  If the original input were
given as "<treeish>^{tree}", there is no confusion.