git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v4 8/8] cat-file: add --batch-command remote-object-info command

From
Calvin Wan <calvinwan@google.com>
Date
May 5, 2022, 18:13 UTC
Message-ID
<CAFySSZCL-ccVK+ZOU5=X4fB3-eu=Cmu-MGpT1cMW+bqmdRcF9w@mail.gmail.com>
In-Reply-To
<20220504212738.162853-1-jonathantanmy@google.com>
> You mention below that this only works in --buffer mode, so I don't
> think that this is relevant.

It works in non-buffer mode as well. I meant that if you wanted to use `remote-object-info` with another option, then --filter is the only one that it works with. I tried to match the documentation for --batch-command that says something similar:

--batch-command=<format> Enter a command mode that reads commands and arguments from stdin. May only be combined with --buffer, --textconv or --filters

> What happens if remote-object-info is mixed with other commands?
It works with other commands. The idea is that you would enter a line like:
remote-object-info <remote> <oid> <oid> ... <oid>
rather than

remote-object-info <remote> <oid> remote-object-info <remote> <oid> remote-object-info <remote> <oid>

since each invocation of remote-object-info would cause another round trip to the server.

> Hmm...I would have expected the information to appear in the same data
> structure used to store other command invocations like "contents",
> "info", and "flush", not separately like this. This design doesn't seem
> to support mixing commands.

Here's an example of how it works with other commands. Let's say I call git cat-file --batch-command --buffer and input the following:

remote-object-info remote1 oid1 oid2 oid3 contents oid4 info oid5 remote-object-info remote2 oid6 oid7 oid8 flush

This would result in the following:

(call transport and store info in separate structure) info oid1 (using passed in object-info) info oid2 info oid3 contents oid4 info oid5 (call transport and store info in separate structure) info oid6 info oid7 info oid8

> Rather than store a pointer like this (and thus needing to keep track of
> the lifetime of the memory pointed to by this pointer), if we only need
> to know whether this queued_cmd is a remote-object-info call, we can
> just use "unsigned is_remote_object_info : 1".
OK
Show 7 quoted lines
>> @@ -565,9 +639,49 @@ static const struct parse_cmd {
>>  } commands[] = {
>>       { "contents", parse_cmd_contents, 1},
>>       { "info", parse_cmd_info, 1},
>> +     { "remote-object-info", parse_cmd_info, 1},
>>       { "flush", NULL, 0},
>>  };
> Double parse_cmd_info?
See above for why remote-object-info becomes info
On Wed, May 4, 2022 at 2:27 PM Jonathan Tan <jonathantanmy@google.com> wrote:
Show 148 quoted lines
>
> Thanks for the patch. I'll limit myself to high-level comments for now,
> since the design may need to change if we want to support mixing
> remote-object-info with other commands in a --batch-command --buffer
> invocation.
>
> Calvin Wan <calvinwan@google.com> writes:
> > Since the `info` command in cat-file --batch-command prints object info
> > for a given object, it is natural to add another command in cat-file
> > --batch-command to print object info for a given object from a remote.
> > Add `remote-object-info` to cat-file --batch-command.
> >
> > While `info` takes object ids one at a time, this creates overhead when
> > making requests to a server so `remote-object-info` instead can take
> > multiple object ids at once.
> >
> > cat-file --batch-command is generally implemented in the following
> > manner:
> >
> >  - Receive and parse input from user
> >  - Call respective function attached to command
> >  - Set batch mode state, get object info, print object info
>
> You mention below that this only works in --buffer mode, so I don't
> think that this is relevant.
>
> > In --buffer mode, this changes to:
> >
> >  - Receive and parse input from user
> >  - Store respective function attached to command in a queue
> >  - After flush, loop through commands in queue
> >     - Call respective function attached to command
> >     - Set batch mode state, get object info, print object info
> >
> > Notice how the getting and printing of object info is accomplished one
> > at a time. As described above, this creates a problem for making
> > requests to a server. Therefore, `remote-object-info` is implemented in
> > the following manner:
>
> If you do want this command to work with --buffer and without, probably
> clearer to replace "Notice" with "Notice that when cat-file is invoked
> both without --buffer and with it". (Having said that, it makes sense to
> me to only make it work with --buffer - making one request per OID would
> probably not be something that the user would want.)
>
> >  - Receive and parse input from user
> >  If command is `remote-object-info`:
> >     - Get object info from remote
> >     - Loop through object info
> >         - Call respective function attached to `info`
> >         - Set batch mode state, use passed in object info, print object
> >           info
> >  Else:
> >     - Call respective function attached to command
> >     - Parse input, get object info, print object info
> >
> > And finally for --buffer mode `remote-object-info`:
> >  - Receive and parse input from user
> >  - Store respective function attached to command in a queue
> >  - After flush, loop through commands in queue:
> >     If command is `remote-object-info`:
> >         - Get object info from remote
> >         - Loop through object info
> >             - Call respective function attached to `info`
> >             - Set batch mode state, use passed in object info, print
> >               object info
> >     Else:
> >         - Call respective function attached to command
> >         - Set batch mode state, get object info, print object info
> >
> > To summarize, `remote-object-info` gets object info from the remote and
> > then generates multiples `info` commands with the object info passed in.
> > It cannot be implemented similarly to `info` and `content` because of
> > the overhead with remote commenication.
>
> What happens if remote-object-info is mixed with other commands?
>
> > diff --git a/Documentation/git-cat-file.txt b/Documentation/git-cat-file.txt
> > index 24a811f0ef..0d9e8e6214 100644
> > --- a/Documentation/git-cat-file.txt
> > +++ b/Documentation/git-cat-file.txt
> > @@ -115,6 +115,10 @@ info <object>::
> >       Print object info for object reference `<object>`. This corresponds to the
> >       output of `--batch-check`.
> >
> > +remote-object-info <remote> [<object>]::
> > +     Print object info for object references `<object>` at specified <remote>.
> > +     This command may only be combined with `--buffer`.
>
> Here you mention that we need --buffer.
>
> > @@ -253,13 +258,14 @@ newline. The available atoms are:
> >
> >  `objectsize:disk`::
> >       The size, in bytes, that the object takes up on disk. See the
> > -     note about on-disk sizes in the `CAVEATS` section below.
> > +     note about on-disk sizes in the `CAVEATS` section below. Not
> > +     supported by `remote-object-info`.
> >
> >  `deltabase`::
> >       If the object is stored as a delta on-disk, this expands to the
> >       full hex representation of the delta base object name.
> >       Otherwise, expands to the null OID (all zeroes). See `CAVEATS`
> > -     below.
> > +     below. Not supported by `remote-object-info`.
>
> OK - makes sense that these are not supported by remote-object-info,
> because the remote may not even store objects in delta format.
>
> > @@ -32,11 +35,14 @@ struct batch_options {
> >       int unordered;
> >       int transform_mode; /* may be 'w' or 'c' for --filters or --textconv */
> >       const char *format;
> > +     int use_remote_info;
> >  };
> >
> >  #define DEFAULT_FORMAT "%(objectname) %(objecttype) %(objectsize)"
> >
> >  static const char *force_path;
> > +static struct object_info *remote_object_info;
> > +static struct oid_array object_info_oids = OID_ARRAY_INIT;
>
> Hmm...I would have expected the information to appear in the same data
> structure used to store other command invocations like "contents",
> "info", and "flush", not separately like this. This design doesn't seem
> to support mixing commands.
>
> > @@ -538,6 +611,7 @@ typedef void (*parse_cmd_fn_t)(struct batch_options *, const char *,
> >  struct queued_cmd {
> >       parse_cmd_fn_t fn;
> >       char *line;
> > +     const char *name;
> >  };
>
> Rather than store a pointer like this (and thus needing to keep track of
> the lifetime of the memory pointed to by this pointer), if we only need
> to know whether this queued_cmd is a remote-object-info call, we can
> just use "unsigned is_remote_object_info : 1".
>
> > @@ -565,9 +639,49 @@ static const struct parse_cmd {
> >  } commands[] = {
> >       { "contents", parse_cmd_contents, 1},
> >       { "info", parse_cmd_info, 1},
> > +     { "remote-object-info", parse_cmd_info, 1},
> >       { "flush", NULL, 0},
> >  };
>
> Double parse_cmd_info?
Previous: Jonathan TanNext: Junio C Hamano
Message 61 of 103 in “fetch —object-info-format: client option for object-info”
  1. fetch —object-info-format: client option for object-infoCalvin Wan, Feb 8, 2022
  2. fetch —object-info-format: client option for object-infoCalvin Wan, Feb 8, 2022
  3. Philip OakleyFeb 9, 2022
  4. Calvin WanFeb 10, 2022
  5. Jonathan TanFeb 9, 2022
  6. Calvin WanFeb 10, 2022
  7. 0/3 object-info: add option for retrieving object infoCalvin Wan, Mar 28, 2022
  8. 1/3 fetch-pack: refactor packet writing and fetch optionsCalvin Wan, Mar 28, 2022
  9. Junio C HamanoMar 29, 2022
  10. Taylor BlauMar 29, 2022
  11. Jonathan TanMar 30, 2022
  12. 2/3 transfer.advertiseObjectInfo: add object-info configCalvin Wan, Mar 28, 2022
  13. Junio C HamanoMar 29, 2022
  14. Calvin WanMar 29, 2022
  15. Taylor BlauMar 29, 2022
  16. 3/3 object-info: add option for retrieving object infoCalvin Wan, Mar 28, 2022
  17. Junio C HamanoMar 29, 2022
  18. Junio C HamanoMar 29, 2022
  19. Junio C HamanoMar 30, 2022
  20. Calvin WanMar 30, 2022
  21. Jonathan TanMar 30, 2022
  22. Junio C HamanoMar 30, 2022
  23. Taylor BlauMar 29, 2022
  24. Calvin WanMar 30, 2022
  25. John CaiMar 30, 2022
  26. Calvin WanMar 31, 2022
  27. Junio C HamanoApr 1, 2022
  28. Jonathan TanMar 30, 2022
  29. Josh SteadmonMar 30, 2022
  30. Calvin WanMar 30, 2022
  31. Junio C HamanoMar 29, 2022
  32. Calvin WanMar 29, 2022
  33. Junio C HamanoMar 31, 2022
  34. 0/8 cat-file: add --batch-command remote-object-info commandCalvin Wan, May 2, 2022
  35. 1/8 fetch-pack: refactor packet writingCalvin Wan, May 2, 2022
  36. 3/8 object-store: add function to free object_info contentsCalvin Wan, May 2, 2022
  37. Junio C HamanoMay 2, 2022
  38. Junio C HamanoMay 4, 2022
  39. Junio C HamanoMay 5, 2022
  40. Calvin WanMay 5, 2022
  41. Junio C HamanoMay 5, 2022
  42. 4/8 object-info: send attribute packet regardless of object idsCalvin Wan, May 2, 2022
  43. Junio C HamanoMay 3, 2022
  44. Calvin WanMay 3, 2022
  45. Jonathan TanMay 3, 2022
  46. 2/8 fetch-pack: move fetch default settingsCalvin Wan, May 2, 2022
  47. Junio C HamanoMay 2, 2022
  48. Jonathan TanMay 3, 2022
  49. Calvin WanMay 5, 2022
  50. 5/8 transport: add client side capability to request object-infoCalvin Wan, May 2, 2022
  51. Junio C HamanoMay 3, 2022
  52. Calvin WanMay 3, 2022
  53. Junio C HamanoMay 4, 2022
  54. Jonathan TanMay 3, 2022
  55. Junio C HamanoMay 4, 2022
  56. 7/8 cat-file: move parse_cmd and DEFAULT_FORMAT upCalvin Wan, May 2, 2022
  57. 6/8 transport: add object-info fallback to fetchCalvin Wan, May 2, 2022
  58. Jonathan TanMay 3, 2022
  59. 8/8 cat-file: add --batch-command remote-object-info commandCalvin Wan, May 2, 2022
  60. Jonathan TanMay 4, 2022
  61. Calvin WanMay 5, 2022
  62. Junio C HamanoMay 5, 2022
  63. Junio C HamanoMay 5, 2022
  64. Calvin WanMay 5, 2022
  65. ZheNing HuJul 31, 2022
  66. Calvin WanAug 8, 2022
  67. 0/6 cat-file: add --batch-command remote-object-info commandCalvin Wan, Jul 28, 2022
  68. Junio C HamanoJul 28, 2022
  69. Junio C HamanoJul 29, 2022
  70. Phillip WoodJul 31, 2022
  71. Calvin WanAug 4, 2022
  72. Junio C HamanoSep 30, 2022
  73. 1/6 fetch-pack: refactor packet writingCalvin Wan, Jul 28, 2022
  74. 2/6 fetch-pack: move fetch initializationCalvin Wan, Jul 28, 2022
  75. 3/6 protocol-caps: initialization bug fixCalvin Wan, Jul 28, 2022
  76. Junio C HamanoJul 29, 2022
  77. 4/6 serve: advertise object-info featureCalvin Wan, Jul 28, 2022
  78. Junio C HamanoJul 29, 2022
  79. Calvin WanAug 1, 2022
  80. Ævar Arnfjörð BjarmasonAug 1, 2022
  81. Junio C HamanoAug 1, 2022
  82. Calvin WanAug 1, 2022
  83. 5/6 transport: add client support for object-infoCalvin Wan, Jul 28, 2022
  84. Junio C HamanoJul 29, 2022
  85. Calvin WanAug 4, 2022
  86. Phillip WoodAug 1, 2022
  87. Calvin WanAug 4, 2022
  88. Phillip WoodAug 8, 2022
  89. Phillip WoodAug 1, 2022
  90. Phillip WoodAug 8, 2022
  91. 6/6 cat-file: add remote-object-info to batch-commandCalvin Wan, Jul 28, 2022
  92. Ævar Arnfjörð BjarmasonJul 29, 2022
  93. ZheNing HuAug 7, 2022
  94. Calvin WanAug 8, 2022
  95. Ævar Arnfjörð BjarmasonAug 11, 2022
  96. Junio C HamanoJul 29, 2022
  97. Calvin WanAug 8, 2022
  98. Ævar Arnfjörð BjarmasonSep 28, 2022
  99. ZheNing HuJul 31, 2022
  100. Calvin WanAug 8, 2022
  101. Junio C HamanoAug 13, 2022
  102. John CaiFeb 9, 2022
  103. Calvin WanFeb 10, 2022

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.