Re: [RFC v2] cat-file: add a --stdin-cmd mode

[John Cai <johncai86@gmail.com>]

Hey Phillip,

First of all thank you for the thorough review—it really helps someone who’s learning how to contribute!

On 27 Jan 2022, at 6:25, Phillip Wood wrote:

> Hi John
>
> On 25/01/2022 22:50, John Cai wrote:
>> This RFC patch proposes a new flag --stdin-cmd that works with
>> git-cat-file --batch. Similar to git-update-ref --stdin, it will accept
>> commands and arguments from stdin.
>>
>> The start of this idea was discussed in [1], where the original
>> motivation was to be able to control when the buffer was flushed to
>> stdout in --buffer mode.
>>
>> However, this can actually be much more useful in situations when
>> git-cat-file --batch is being used as a long lived backend query
>> process. At GitLab, we use a pair of cat-file processes. One for
>> iterating over object metadata with --batch-check, and the other to grab
>> object contents with --batch. However, if we had --stdin-cmd, we could
>> get rid of the second --batch-check process, and just have one progress
>> where we can flip between getting object info, and getting object contents.
>> This can lead to huge savings.
>>
>> git cat-file --batch --stdin-cmd
>>
>> $ <command> [arg1] [arg2] NL
>>
>> We can also add a -z mode to allow for NUL-terminated lines
>>
>> $ <command> [arg1] [arg2] NUL
>>
>> This patch adds three commands: object, info, fflush
>>
>> $ object <sha1> NL
>> $ info <sha1> NL
>> $ fflush NL
>>
>> These three would be immediately useful in GitLab's context, but one can
>> imagine this mode to be further extended for other things.
>>
>> For instance, a non-trivial part of "cat-file --batch" time is spent
>> on parsing its argument and seeing if it's a revision, ref etc. So we
>> could add a command that only accepts a full-length 40
>> character SHA-1.
>>
>> This would be the first step in adding such an interface to
>> git-cat-file.
>>
>> [1] https://lore.kernel.org/git/pull.1124.git.git.1636149400.gitgitgadget@gmail.com/
>>
>> Helped-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
>> Signed-off-by: John Cai <johncai86@gmail.com>
>> ---
>> Changes from v1:
>>
>> - changed option name to batch-command.
>> - changed command function interface to receive the whole line after the command
>>    name to put the onus of parsing arguments to each individual command function.
>> - pass in whole line to batch_one_object in both parse_cmd_object and
>>    parse_cmd_info to support spaces in the object reference.
>> - removed addition of -z to include in a separate patch series
>> - added documentation.
>
> I've left some comments below, they're mostly small details, I like the new option name and the changes you've made to the command parsing.
>
>> ---
>>   Documentation/git-cat-file.txt |  15 +++++
>>   builtin/cat-file.c             | 114 ++++++++++++++++++++++++++++++++-
>>   strvec.c                       |  23 +++++++
>>   strvec.h                       |   8 +++
>>   t/t1006-cat-file.sh            |  32 +++++++++
>>   5 files changed, 191 insertions(+), 1 deletion(-)
>>
>> diff --git a/Documentation/git-cat-file.txt b/Documentation/git-cat-file.txt
>> index bef76f4dd0..8aefa45e4c 100644
>> --- a/Documentation/git-cat-file.txt
>> +++ b/Documentation/git-cat-file.txt
>> @@ -96,6 +96,21 @@ OPTIONS
>>   	need to specify the path, separated by whitespace.  See the
>>   	section `BATCH OUTPUT` below for details.
>>  +-batch-command::
>
> is this missing a '-'?
>
>> +	Enter a command mode that reads from stdin. May not be combined with any
>> +	other options or arguments except `--textconv` or `--filters`, in which
>> +	case the input lines also need to specify the path, separated by
>> +	whitespace.  See the section `BATCH OUTPUT` below for details.
>> +
>> +object <object>::
>> +	Print object contents for object reference <object>
>> +
>> +info <object>::
>> +	Print object info for object reference <object>
>> +
>> +flush::
>> +	Flush to stdout immediately when used with --buffer
>> +
>>   --batch-all-objects::
>>   	Instead of reading a list of objects on stdin, perform the
>>   	requested batch operation on all objects in the repository and
>> diff --git a/builtin/cat-file.c b/builtin/cat-file.c
>> index 7b3f42950e..30794284d5 100644
>> --- a/builtin/cat-file.c
>> +++ b/builtin/cat-file.c
>> @@ -16,6 +16,7 @@
>>   #include "packfile.h"
>>   #include "object-store.h"
>>   #include "promisor-remote.h"
>> +#include "strvec.h"
>>    struct batch_options {
>>   	int enabled;
>> @@ -26,7 +27,10 @@ struct batch_options {
>>   	int unordered;
>>   	int cmdmode; /* may be 'w' or 'c' for --filters or --textconv */
>>   	const char *format;
>> +	int stdin_cmd;
>
> Now that the option has been renamed it would be nice to rename the corresponding variable to match

I was trying to find a good name. There is already a cmdmode variable. I’m thinking stdin_cmd is not such a
bad name since we are receiving commands from stdin. suggestions welcome!

>
>> +	int end_null;
>
> If you're not adding '-z' here then you don't need this or the addition below.
>
>>   };
>> +static char line_termination = '\n';
>>    static const char *force_path;
>>  @@ -508,6 +512,102 @@ static int batch_unordered_packed(const struct object_id *oid,
>>   				      data);
>>   }
>>  +enum batch_state {
>> +	/* Non-transactional state open for commands. */
>> +	BATCH_STATE_OPEN,
>> +};
>
> I forgot to ask what the idea behind the batch state is last time - what's it for?

This is to support transactional semantics for commands we want to support in the future, but
since this is already a biggish change, we can leave this out of this series.

>
>> +static void parse_cmd_object(struct batch_options *opt,
>> +			     const char *line,
>> +			     struct strbuf *output,
>> +			     struct expand_data *data)
>> +{
>> +	opt->print_contents = 1;
>> +	batch_one_object(line, output, opt, data);
>> +}
>> +
>> +static void parse_cmd_info(struct batch_options *opt,
>> +			   const char *line,
>> +			   struct strbuf *output,
>> +			   struct expand_data *data)
>> +{
>> +	opt->print_contents = 0;
>> +	batch_one_object(line, output, opt, data);
>> +}
>> +
>> +static void parse_cmd_fflush(struct batch_options *opt,
>> +			     const char *line,
>> +			     struct strbuf *output,
>> +			     struct expand_data *data)
>> +{
>> +	fflush(stdout);
>> +}
>> +
>> +typedef void (*parse_cmd_fn_t)(struct batch_options *, const char *,
>> +			       struct strbuf *, struct expand_data *);
>> +
>> +static const struct parse_cmd {
>> +	const char *prefix;
>> +	parse_cmd_fn_t fn;
>> +	unsigned args;
>
> This is now a flag so maybe 'takes_args' would better describe its purpose.
>
>> +	enum batch_state state;
>> +} commands[] = {
>> +	{ "object", parse_cmd_object, 1, BATCH_STATE_OPEN },
>> +	{ "info", parse_cmd_info, 1, BATCH_STATE_OPEN },
>> +	{ "fflush", parse_cmd_fflush, 0, BATCH_STATE_OPEN },
>> +};
>> +
>> +static void batch_objects_stdin_cmd(struct batch_options *opt,
>> +				    struct strbuf *output,
>> +				    struct expand_data *data)
>> +{
>> +	struct strbuf input = STRBUF_INIT;
>> +	enum batch_state state = BATCH_STATE_OPEN;
>> +
>> +	/* Read each line dispatch its command */
>> +	while (!strbuf_getwholeline(&input, stdin, line_termination)) {
>> +		int i;
>> +		const struct parse_cmd *cmd = NULL;
>> +		const char *p;
>> +
>> +		if (*input.buf == line_termination)
>> +			die("empty command in input");
>> +		else if (isspace(*input.buf))
>> +			die("whitespace before command: %s", input.buf);
>> +
>> +		for (i = 0; i < ARRAY_SIZE(commands); i++) {
>> +			const char *prefix = commands[i].prefix;
>> +			char c;
>> +			const char *cmd_name;
>
> skip_prefix() sets this to the end of the name so maybe 'cmd_end' would be clearer?
>
>> +			if (!skip_prefix(input.buf, prefix, &cmd_name))
>> +				continue;
>> +			/*
>> +			 * If the command has arguments, verify that it's
>> +			 * followed by a space. Otherwise, it shall be followed
>> +			 * by a line terminator.
>> +			 */
>> +			c = commands[i].args ? ' ' : line_termination;
>> +			if (input.buf[strlen(prefix)] != c)
>
> Now that you're using skip_prefix() you can write this as
>     		if (*cmd_end != c)
>
>> +				continue;
>> +
>> +			cmd = &commands[i];
>> +			break;
>> +		}
>> +		if (!cmd)
>> +			die("unknown command: %s", input.buf);
>> +
>> +		p = input.buf + strlen(cmd->prefix) + 1;
>
> This can be simplified to
>     	p = cmd_end + 1;
>
>> +		const char *pos = strstr(p, &line_termination);
>
> This isn't needed without '-z'. If it were required then using strchrnul() would prevent a NULL pointer dereference when the last input line does not end with a terminator. I think we typically call a pointer to the end of the line 'eol' or 'end'. Also variables should be declared at the top of the function.
>
>> +		switch (state) {
>> +		case BATCH_STATE_OPEN:
>> +			break;
>> +		}
>> +		cmd->fn(opt, xstrndup(p, pos-p), output, data);
>
> Is there a reason this is passing a copy of the string?
>
>> +	}
>> +	strbuf_release(&input);
>> +}
>> +
>>   static int batch_objects(struct batch_options *opt)
>>   {
>>   	struct strbuf input = STRBUF_INIT;
>> @@ -515,6 +615,7 @@ static int batch_objects(struct batch_options *opt)
>>   	struct expand_data data;
>>   	int save_warning;
>>   	int retval = 0;
>> +	const int stdin_cmd = opt->stdin_cmd;
>>    	if (!opt->format)
>>   		opt->format = "%(objectname) %(objecttype) %(objectsize)";
>> @@ -590,7 +691,8 @@ static int batch_objects(struct batch_options *opt)
>>   	save_warning = warn_on_object_refname_ambiguity;
>>   	warn_on_object_refname_ambiguity = 0;
>>  -	while (strbuf_getline(&input, stdin) != EOF) {
>> +	while (!stdin_cmd &&
>
> If you moved the 'if (stdin_cmd)' block above this block we could loose this change. I'm not sure if that is possible without looking at the whole function though.
>
>> +	       strbuf_getline(&input, stdin) != EOF) {
>>   		if (data.split_on_whitespace) {
>>   			/*
>>   			 * Split at first whitespace, tying off the beginning
>> @@ -608,6 +710,9 @@ static int batch_objects(struct batch_options *opt)
>>   		batch_one_object(input.buf, &output, opt, &data);
>>   	}
>>  +	if (stdin_cmd)
>> +		batch_objects_stdin_cmd(opt, &output, &data);
>> +
>>   	strbuf_release(&input);
>>   	strbuf_release(&output);
>>   	warn_on_object_refname_ambiguity = save_warning;
>> @@ -636,6 +741,7 @@ static int batch_option_callback(const struct option *opt,
>>    	bo->enabled = 1;
>>   	bo->print_contents = !strcmp(opt->long_name, "batch");
>> +	bo->stdin_cmd = !strcmp(opt->long_name, "batch-command");
>>   	bo->format = arg;
>>    	return 0;
>> @@ -683,6 +789,10 @@ int cmd_cat_file(int argc, const char **argv, const char *prefix)
>>   			N_("like --batch, but don't emit <contents>"),
>>   			PARSE_OPT_OPTARG | PARSE_OPT_NONEG,
>>   			batch_option_callback),
>> +		OPT_CALLBACK_F(0, "batch-command", &batch, N_(""),
>> +			 N_("enters batch mode that accepts commands"),
>> +			 PARSE_OPT_NOARG | PARSE_OPT_NONEG,
>> +			 batch_option_callback),
>>   		OPT_CMDMODE(0, "batch-all-objects", &opt,
>>   			    N_("with --batch[-check]: ignores stdin, batches all known objects"), 'b'),
>>   		/* Batch-specific options */
>> @@ -738,6 +848,8 @@ int cmd_cat_file(int argc, const char **argv, const char *prefix)
>>   	/* Batch defaults */
>>   	if (batch.buffer_output < 0)
>>   		batch.buffer_output = batch.all_objects;
>> +	if (batch.end_null)
>> +		line_termination = '\0';
>>    	/* Return early if we're in batch mode? */
>>   	if (batch.enabled) {
>> diff --git a/strvec.c b/strvec.c
>> index 61a76ce6cb..7dca04bf7a 100644
>> --- a/strvec.c
>> +++ b/strvec.c
>> [...]
>
> We don't need any strvec changes now that we don't split the input lines to --bactch-command
>
>> +F='%s\0'
>
> This isn't used now
>
>> +test_expect_success 'batch-command unknown command' '
>> +	echo unknown_command >cmd &&
>> +	test_expect_code 128 git cat-file --batch-command < cmd 2>err &&
>> +	grep -E "^fatal:.*unknown command.*" err
>> +'
>> +
>> +test_expect_success 'setup object data' '
>> +	content="Object Data" &&
>> +	size=$(strlen "$content") &&
>> +	sha1=$(echo_without_newline "$content" | git hash-object -w --stdin)
>> +'
>> +
>> +test_expect_success 'batch-command calling object works' '
>> +	echo "object $sha1" | git cat-file --batch-command >actual &&
>> +	echo "$sha1 blob $size" >expect &&
>> +	echo `git cat-file -p "$sha1"` >>expect &&
>> +	test_cmp expect actual
>> +'
>> +
>> +test_expect_success 'batch-command calling info works' '
>> +	echo "info $sha1" | git cat-file --batch-command >actual &&
>> +	echo "$sha1 blob $size" >expect &&
>> +	test_cmp expect actual
>> +'
>
> I had a quick look at this test file and there is a loop at the top that runs some --batch tests on various inputs, I wonder if these two tests could go in there.
>
>> +test_expect_success 'batch-command fflush works' '
>> +	printf "fflush\n" > cmd &&
>> +	test_expect_code 0 git cat-file --batch-command < cmd 2>err
>> +'
>
> It'd be nice to check this actually flushes the output.

could you give me some ideas on how to do this?

>
> Best Wishes
>
> Phillip
>
>>   test_done