Re: [PATCH 4/4] btrfs-progs: change -t option for subvolume list to print a simple space-separated table (making it machine-readable)

[Axel Burri <axel@tty0.ch>]

On 2015-10-05 17:42, Goffredo Baroncelli wrote:
> Hi Axel,
> 
> On 2015-10-05 17:04, Axel Burri wrote:
> [...]
>> I still don't think it is wise to bloat things further just for printing
>> nice tables. My impression is that "btrfs subvolume list" is
>> human-readable enough without the '-t' flag, while the output with '-t'
>> flag is much more machine-readable-friendly, and thus should have the
>> highest possible performance. e.g.:
> 
> I disagree, the "-t" mode is for the human; the non '-t' mode would be for machine (if you take the space as separator and the first word as key, with the exception of the path which starts from the "path " word and ends to the end of line); unfortunately "top level" is an exception. Anyway btrfs is not very machine-friendly (sic).

That's what my patches are all about in the first place, to make it more
machine-friendly :)

> Compare the two output below:
> 
> $ sudo btrfs sub list -a /
> ID 257 gen 44309 top level 5 path <FS_TREE>/debian
> ID 289 gen 17415 top level 257 path debian/var/lib/machines
> ID 298 gen 35434 top level 5 path <FS_TREE>/debian-i32
> ID 299 gen 43961 top level 5 path <FS_TREE>/boot
> ID 300 gen 39452 top level 5 path <FS_TREE>/debian-jessie

"grep-friendly"

> 
> $ sudo btrfs sub list -at / 
> ID	gen	top level	path	
> --	---	---------	----	
> 257	44310	5		<FS_TREE>/debian
> 289	17415	257		debian/var/lib/machines
> 298	35434	5		<FS_TREE>/debian-i32
> 299	43961	5		<FS_TREE>/boot
> 300	39452	5		<FS_TREE>/debian-jessie
> 
> 
> I think that is easy to declare the latter more "human" friendly. 

Well yes, but in some ways the latter is also more machine-friendly, as
there is less data to pipe and a whitespace-separated list is much
easier to parse (of course the same applies to your table implementation
below).

Problems here are:

- "top level" (whitespace, addressed in [PATCH 4/4])

- timestamp (-s) "YYYY-MM-DD HH:MM:SS" (whitspace and no timezone,
addressed in [PATCH 3/4])

- path can contain whitespace (not _that_ important as it's always
printed last).

> 
> BTW with the use of the table_* function the output become:
> 
> $ sudo ./btrfs sub list -at / 
> ID  gen   top level path                   
> === ===== ========= =======================
> 257 44311         5 <FS_TREE>/debian       
> 289 17415       257 debian/var/lib/machines
> 298 35434         5 <FS_TREE>/debian-i32   
> 299 43961         5 <FS_TREE>/boot         
> 300 39452         5 <FS_TREE>/debian-jessie
> 
> $ sudo ./btrfs sub list -aptcguqR / 
> ID  gen   cgen  parent top level parent_uuid received_uuid uuid                                 path                   
> === ===== ===== ====== ========= =========== ============= ==================================== =======================
> 257 44313     7      5         5           -             - 840c86cf-e78b-d54a-ab38-66662858812d <FS_TREE>/debian       
> 289 17415 17415    257       257           -             - 8b857250-3a3e-754d-810e-57342bbb2f56 debian/var/lib/machines
> 298 35434 35399      5         5           -             - 1f38049b-b153-d741-b903-d2de6fd7b3fd <FS_TREE>/debian-i32   
> 299 43961 35512      5         5           -             - f9d52b6b-a6d1-8c45-a6cd-ddb68cf58062 <FS_TREE>/boot         
> 300 39452 37744      5         5           -             - 026e44bd-66d4-e341-9a14-0124acf79793 <FS_TREE>/debian-jessie
> 
> I right aligned each field with the exception of the path (which is left aligned).
> I already prepared the patch which convert "btrfs sub list -t" to use the table_* function. If you want I can send it to you to extend/replace your patch #4.

Looks promising, although I'd still not recommend it because of the
increased memory/cpu requirements.