mbox series

[0/5] promise: introduce promises to track success or error

Message ID pull.1666.git.git.1708241612.gitgitgadget@gmail.com (mailing list archive)
Headers show
Series promise: introduce promises to track success or error | expand

Message

Philippe Blain via GitGitGadget Feb. 18, 2024, 7:33 a.m. UTC
Hello all, this is my first patchset so thank you for being patient with me
as I learn the process and conventions of your very fine project. These
patches are intended as part of the Libification effort, to show how we
could use a Promise structure to help return values from functions.


Problems
========

We seek to make libification easier by establishing a pattern for tracking
whether a function errored in a rich way. Currently, any given function
could immediately die(), or use error() to print directly to the console,
bypassing any relevant verbosity checks. The use of die() currently makes
use of Git as a library inconvenient since it is not graceful.

Additionally, returning using return error(...) (as is commonly done) always
just returns a generic error value, -1, which provides little information.


Approach
========

I solve this problem by splitting the single return value into two return
values: error, and message. However, managing two output variables can
require some coordination, and this coordination can be abstracted away by
use of an existing pattern named Promise.


Promise Concept
===============

A promise is a contract representing "some task" that will eventually
complete. Initially a promise is considered in a pending state. When it
completes, one of two codepaths will eventually be entered: reject, or
resolve. Once resolved or rejected, the promise enters a different state
representing the result. Reject or resolve may only be called once on a
given promise.

Until now, everything I described up to this point is consistent with other
implementations, such as the ECMAScript standard for promises. However, this
implementation departs from the complexity of those promises. In this
implementation, promises are simple and canNOT be chained using .then(...)
and do NOT have any notion of automatic bubbling (via re-entering the
pending state).


Sample output and reproduction
==============================

During an error, we can have richer feedback as to what caused the problem.

% git apply garbage.patch
error: 
    could not find header
caused by:
    patch fragment without header at line 1: @@ -2 +2 @@


To reproduce this output, you can use the following patch (garbage.patch):

@@ -2 +2 @@



Goals
=====

I would love to get feedback on this approach. This patchset is kept small,
so as to serve as a minimal proof of concept. It is intended to abstract to
asynchronous use-cases even though this is only a synchronous one.
Eventually, any top-level function, such as apply_all_patches(...) would
return its output via a promise to make the library interface as clean as
possible, but this patchset does not accomplish this goal. Hopefully it can
provide a direction to go in to achieve that.


Diversion
=========

While building this patchset, I noted a bug that may not have a concrete
repro case in the master branch. The bug is that when invoking git am, it
can call out to git apply, passing many flags but interestingly not the
--quiet flag. I included a fix for this issue in the patchset.


Potential Issue
===============

There is one difficulty with this approach, which is the high level of
repetition in the code required. Tracking which promise is which is its own
source of complexity and may make mistakes more prone to happen. If anyone
has suggestions for how to make the code cleaner, I would love to hear.

Thank you, Philip

Philip Peterson (5):
  promise: add promise pattern to track success/error from operations
  apply: use new promise structures in git-apply logic as a proving
    ground
  apply: update t4012 test suite
  apply: pass through quiet flag to fix t4150
  am: update test t4254 by adding the new error text

 Makefile               |   1 +
 apply.c                | 133 +++++++++++++++++++++++++++--------------
 apply.h                |   9 ++-
 builtin/am.c           |   5 ++
 promise.c              |  89 +++++++++++++++++++++++++++
 promise.h              |  71 ++++++++++++++++++++++
 range-diff.c           |  14 +++--
 t/t4012-diff-binary.sh |   4 +-
 t/t4254-am-corrupt.sh  |   9 ++-
 9 files changed, 279 insertions(+), 56 deletions(-)
 create mode 100644 promise.c
 create mode 100644 promise.h


base-commit: 2996f11c1d11ab68823f0939b6469dedc2b9ab90
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1666%2Fphilip-peterson%2Fpeterson%2Femail-v1
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1666/philip-peterson/peterson/email-v1
Pull-Request: https://github.com/git/git/pull/1666

Comments

Phillip Wood Feb. 19, 2024, 2:25 p.m. UTC | #1
Hi Philip

On 18/02/2024 07:33, Philip Peterson via GitGitGadget wrote:
> Hello all, this is my first patchset so thank you for being patient with me
> as I learn the process and conventions of your very fine project. These
> patches are intended as part of the Libification effort, to show how we
> could use a Promise structure to help return values from functions.

I agree that we could do with a better way of propagating errors ups the 
call-chain that (a) allows us to pass more detailed information about 
the error to the caller and (b) add useful context to the error message 
as the stack in unwound. I'm afraid I do not think that the promise 
implementation in this patch series is a good way forward for several 
reasons.

1) It is hard to see how we can wrap the return value of the function in 
a promise and preserve type safety. Even if we used some  kind of union 
the compiler will not warn us if the caller reads a different member to 
the one that the callee set.

2) It obscures the return type of the function and forces callers to 
unpack the value from the promise adding complexity to the calling code.

3) It imposes a cost in terms of dynamic memory allocation on code that 
is called synchronously and therefore does not need to allocate the 
promise on the heap. This adds complexity and is sure to result in 
memory leaks.

4) If the function fails we need to propagate the error using 
PROMISE_BUBBLE_UP() which forces the caller to add more context to the 
error message even if it does covey anything useful to the user. For 
example in patch 5 we see

     error:
	could not find header
     caused by:
	could not find file diff header
     caused by:
	git diff header lacks filename information (line 4)" >expected

The error message starts by saying it couldn't find the header and ends 
by saying it did actually find the header but it could not parse it.

5) The cover letter talks about adding asynchronous examples in the 
future but it is not clear what part of the git code base it is 
referring to.

I think we'd be better served by some kind of structured error type like 
the failure_result in this patch series that is allocated on the stack 
by the caller at the entry point to the library and passed down the call 
chain. That avoids the need for lots of dynamic allocations and allows 
us to continue allocating "out" parameters on the stack. For example

     int f(struct repository *r) {
	struct object_id oid;

	if (repo_get_oid(r, "HEAD", &oid))
		return error(_("could not parse HEAD"))

	/* use oid here */
     }

would become
     int f(struct repository *r, struct error *err) {
	struct object_id oid;

	if (repo_get_oid(r, "HEAD", &oid))
		return error(&err, _("could not parse HEAD"))

	/* use oid here */
     }

I'm sure this has been discussed in the past but I didn't manage to turn 
anything up with a quick search of the archive on lore.kernel.org.

Best Wishes

Phillip



> Problems
> ========
> 
> We seek to make libification easier by establishing a pattern for tracking
> whether a function errored in a rich way. Currently, any given function
> could immediately die(), or use error() to print directly to the console,
> bypassing any relevant verbosity checks. The use of die() currently makes
> use of Git as a library inconvenient since it is not graceful.
> 
> Additionally, returning using return error(...) (as is commonly done) always
> just returns a generic error value, -1, which provides little information.
> 
> 
> Approach
> ========
> 
> I solve this problem by splitting the single return value into two return
> values: error, and message. However, managing two output variables can
> require some coordination, and this coordination can be abstracted away by
> use of an existing pattern named Promise.
> 
> 
> Promise Concept
> ===============
> 
> A promise is a contract representing "some task" that will eventually
> complete. Initially a promise is considered in a pending state. When it
> completes, one of two codepaths will eventually be entered: reject, or
> resolve. Once resolved or rejected, the promise enters a different state
> representing the result. Reject or resolve may only be called once on a
> given promise.
> 
> Until now, everything I described up to this point is consistent with other
> implementations, such as the ECMAScript standard for promises. However, this
> implementation departs from the complexity of those promises. In this
> implementation, promises are simple and canNOT be chained using .then(...)
> and do NOT have any notion of automatic bubbling (via re-entering the
> pending state).
> 
> 
> Sample output and reproduction
> ==============================
> 
> During an error, we can have richer feedback as to what caused the problem.
> 
> % git apply garbage.patch
> error:
>      could not find header
> caused by:
>      patch fragment without header at line 1: @@ -2 +2 @@
> 
> 
> To reproduce this output, you can use the following patch (garbage.patch):
> 
> @@ -2 +2 @@
> 
> 
> 
> Goals
> =====
> 
> I would love to get feedback on this approach. This patchset is kept small,
> so as to serve as a minimal proof of concept. It is intended to abstract to
> asynchronous use-cases even though this is only a synchronous one.
> Eventually, any top-level function, such as apply_all_patches(...) would
> return its output via a promise to make the library interface as clean as
> possible, but this patchset does not accomplish this goal. Hopefully it can
> provide a direction to go in to achieve that.
> 
> 
> Diversion
> =========
> 
> While building this patchset, I noted a bug that may not have a concrete
> repro case in the master branch. The bug is that when invoking git am, it
> can call out to git apply, passing many flags but interestingly not the
> --quiet flag. I included a fix for this issue in the patchset.
> 
> 
> Potential Issue
> ===============
> 
> There is one difficulty with this approach, which is the high level of
> repetition in the code required. Tracking which promise is which is its own
> source of complexity and may make mistakes more prone to happen. If anyone
> has suggestions for how to make the code cleaner, I would love to hear.
> 
> Thank you, Philip
> 
> Philip Peterson (5):
>    promise: add promise pattern to track success/error from operations
>    apply: use new promise structures in git-apply logic as a proving
>      ground
>    apply: update t4012 test suite
>    apply: pass through quiet flag to fix t4150
>    am: update test t4254 by adding the new error text
> 
>   Makefile               |   1 +
>   apply.c                | 133 +++++++++++++++++++++++++++--------------
>   apply.h                |   9 ++-
>   builtin/am.c           |   5 ++
>   promise.c              |  89 +++++++++++++++++++++++++++
>   promise.h              |  71 ++++++++++++++++++++++
>   range-diff.c           |  14 +++--
>   t/t4012-diff-binary.sh |   4 +-
>   t/t4254-am-corrupt.sh  |   9 ++-
>   9 files changed, 279 insertions(+), 56 deletions(-)
>   create mode 100644 promise.c
>   create mode 100644 promise.h
> 
> 
> base-commit: 2996f11c1d11ab68823f0939b6469dedc2b9ab90
> Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1666%2Fphilip-peterson%2Fpeterson%2Femail-v1
> Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1666/philip-peterson/peterson/email-v1
> Pull-Request: https://github.com/git/git/pull/1666
Jeff King Feb. 20, 2024, 2:57 a.m. UTC | #2
On Mon, Feb 19, 2024 at 02:25:29PM +0000, Phillip Wood wrote:

> I think we'd be better served by some kind of structured error type like the
> failure_result in this patch series that is allocated on the stack by the
> caller at the entry point to the library and passed down the call chain.
> That avoids the need for lots of dynamic allocations and allows us to
> continue allocating "out" parameters on the stack. For example
> 
>     int f(struct repository *r) {
> 	struct object_id oid;
> 
> 	if (repo_get_oid(r, "HEAD", &oid))
> 		return error(_("could not parse HEAD"))
> 
> 	/* use oid here */
>     }
> 
> would become
>     int f(struct repository *r, struct error *err) {
> 	struct object_id oid;
> 
> 	if (repo_get_oid(r, "HEAD", &oid))
> 		return error(&err, _("could not parse HEAD"))
> 
> 	/* use oid here */
>     }
> 
> I'm sure this has been discussed in the past but I didn't manage to turn
> anything up with a quick search of the archive on lore.kernel.org.

There's some discussion in this sub-thread:

  https://lore.kernel.org/git/20171103191309.sth4zjokgcupvk2e@sigill.intra.peff.net/

that also references this earlier thread:

  https://lore.kernel.org/git/20160927191955.mympqgylrxhkp24n@sigill.intra.peff.net/

I still think this is a reasonable way to go. At one point I had a
proof-of-concept conversion of some of the ref code, but I don't think I
have it any more.

-Peff
Phillip Wood Feb. 20, 2024, 10:53 a.m. UTC | #3
Hi Peff

On 20/02/2024 02:57, Jeff King wrote:
> On Mon, Feb 19, 2024 at 02:25:29PM +0000, Phillip Wood wrote:
> 
>> I'm sure this has been discussed in the past but I didn't manage to turn
>> anything up with a quick search of the archive on lore.kernel.org.
> 
> There's some discussion in this sub-thread:
> 
>    https://lore.kernel.org/git/20171103191309.sth4zjokgcupvk2e@sigill.intra.peff.net/
> 
> that also references this earlier thread:
> 
>    https://lore.kernel.org/git/20160927191955.mympqgylrxhkp24n@sigill.intra.peff.net/

Thanks for digging up those links

> I still think this is a reasonable way to go. At one point I had a
> proof-of-concept conversion of some of the ref code, but I don't think I
> have it any more.

Ah, that's interesting - the ref transaction functions already take a 
struct strbuf to populate with an error message so maybe that would be a 
simple place to start with a conversion to an error struct.

Best Wishes

Phillip
Patrick Steinhardt Feb. 20, 2024, 12:19 p.m. UTC | #4
On Tue, Feb 20, 2024 at 10:53:06AM +0000, Phillip Wood wrote:
> Hi Peff
> 
> On 20/02/2024 02:57, Jeff King wrote:
> > On Mon, Feb 19, 2024 at 02:25:29PM +0000, Phillip Wood wrote:
> > 
> > > I'm sure this has been discussed in the past but I didn't manage to turn
> > > anything up with a quick search of the archive on lore.kernel.org.
> > 
> > There's some discussion in this sub-thread:
> > 
> >    https://lore.kernel.org/git/20171103191309.sth4zjokgcupvk2e@sigill.intra.peff.net/
> > 
> > that also references this earlier thread:
> > 
> >    https://lore.kernel.org/git/20160927191955.mympqgylrxhkp24n@sigill.intra.peff.net/
> 
> Thanks for digging up those links
> 
> > I still think this is a reasonable way to go. At one point I had a
> > proof-of-concept conversion of some of the ref code, but I don't think I
> > have it any more.
> 
> Ah, that's interesting - the ref transaction functions already take a struct
> strbuf to populate with an error message so maybe that would be a simple
> place to start with a conversion to an error struct.

I would certainly welcome such a change. It might make sense to wait a
bit until the reftable dust has settled, but once that code has landed
I'd be quite happy to see improvements to our error handling.

While we're already at it throwing ideas around, I also have to wonder
whether this would be a long-term solution towards computer-friendly
errors. One of the problems we quite frequently hit in Gitaly is that we
are forced to parse error messages in order to figure out why exactly
something has failed. Needless to say, this is quite fragile and also
feels very wrong.

Now if we had a more structured way to pass errors around this might
also enable us to convey more meaning to the caller of Git commands. In
a hypothetical world where all errors were using an explicit error type,
then this error type could eventually become richer and contain more
information that is relevant to the calling script. And if such rich
error information was available, then it would not be that far fetched
to ask Git to emit errors in a computer-parsable format like for example
JSON.

Patrick
Junio C Hamano Feb. 20, 2024, 4:20 p.m. UTC | #5
Patrick Steinhardt <ps@pks.im> writes:

> While we're already at it throwing ideas around, I also have to wonder
> whether this would be a long-term solution towards computer-friendly
> errors. One of the problems we quite frequently hit in Gitaly is that we
> are forced to parse error messages in order to figure out why exactly
> something has failed. Needless to say, this is quite fragile and also
> feels very wrong.
>
> Now if we had a more structured way to pass errors around this might
> also enable us to convey more meaning to the caller of Git commands. In
> a hypothetical world where all errors were using an explicit error type,
> then this error type could eventually become richer and contain more
> information that is relevant to the calling script. And if such rich
> error information was available, then it would not be that far fetched
> to ask Git to emit errors in a computer-parsable format like for example
> JSON.

I do not know about the "JSON-parseable" part, but a structured
error message, or even just a set of error codes that can be
recorded in an index, might already be a great improvement.
Jeff King Feb. 21, 2024, 6:03 p.m. UTC | #6
On Tue, Feb 20, 2024 at 01:19:16PM +0100, Patrick Steinhardt wrote:

> While we're already at it throwing ideas around, I also have to wonder
> whether this would be a long-term solution towards computer-friendly
> errors. One of the problems we quite frequently hit in Gitaly is that we
> are forced to parse error messages in order to figure out why exactly
> something has failed. Needless to say, this is quite fragile and also
> feels very wrong.
> 
> Now if we had a more structured way to pass errors around this might
> also enable us to convey more meaning to the caller of Git commands. In
> a hypothetical world where all errors were using an explicit error type,
> then this error type could eventually become richer and contain more
> information that is relevant to the calling script. And if such rich
> error information was available, then it would not be that far fetched
> to ask Git to emit errors in a computer-parsable format like for example
> JSON.

I think what I'm proposing (and if I understand correctly what Phillip
was thinking) is somewhat orthogonal. I agree that structured errors are
nice for computers to read. But it does open up a pretty big can of
worms regarding classifying each error, especially as root causes are
often multi-level.

For example, imagine that the caller asks to resolve a ref. We might get
a syscall error opening the loose ref. Or we might get one opening the
packed-refs file (in a reftable world, you might imagine errors opening
various other files). What is the structured error? Obviously it is "we
can't resolve that ref" at some level. But the caller might want to know
about the failed open (whether it is just ENOENT, or if we ran into
EPERM or even EIO).

Or looking at higher levels; if I ask for the merge base between A and
B, but one of those can't be resolved, how do we communicate that error?
It is some sort of "cannot resolve" error, but it needs to be
parameterized to know which is which.

All of those questions can be answered, of course, but now we are
developing a micro-format that lets us describe all errors in a
standardized way. And I think that is going to put a burden on the code
which is generating the errors (and potentially on the consumers, too,
if they have to decipher the structure to figure out what they want).

Whereas what I was really advocating for is punting on the structured
thing entirely. We keep our unstructured string errors for the most
part, but we simply let the caller pass in a context that tells us what
to do with them. That lets us keep providing specific error messages
from low-level functions without printing to stderr or exiting, which
higher-level code (especially lib-ified code) would not want.

I think it could also be the first building block for making more
structured errors (since those low-level callers are free to provide
lots of details), but it doesn't have to be.

-Peff
Philip Peterson March 12, 2024, 4:18 a.m. UTC | #7
Hi all,

Thanks for the feedback. Reading between the lines, it sounds like we are not
quite happy with the flavor of the approach. To deconstruct, so we can evaluate
each axis:

   1. Canonical indication of pass/fail (as opposed to implicit rules about
      return value being negative or positive or nonzero indicating failure)

   2. Coupling of control flow with result reporting Bubble-up functionality -
      Essentially a convention for handling die()-like functionality without
      calling die() and ending the process

   3. Further error context via breadcrumbs (Foo errored because Bar errored
      because Baz errored)

   4. Error codes - Rather than just -1, having a richer, named space of errors

   5. Rich error data - void *data attached to each error (not actually proposed
      as part of promise but may be relevant)

   6. Async-capable - shelving this since it may require further justification.

First, bubble-up functionality (#3) seems like it will be critical to support if
we intend to stop calling die(). This can be done either exhaustively (by
manually checking each error to see if it is fatal, and if so, invoking a
control-flow action), or by a macro. I couldn’t find any established conventions
for doing this in C, and the original patch series doesn’t really address this
well either since it requires a manual `if (...)` check, so it is an area for
further exploration. (Actually another proposed alternative was `longjmp`, but
it sounds like that is a can of worms to be avoided.)

Back to the overall approach. To contrast to one alternative, error reporting
functions a la the intriguing thread by Peff:

https://lore.kernel.org/git/20160927191955.mympqgylrxhkp24n@sigill.intra.peff.net/

With that pattern, there is no concept of an overall status, either pass or
fail. One must assume that if there was an error, then the operation failed, and
if no error, then it passed. But then there is a question of whether an error
was fatal. We know that some errors are fatal (meaning an immediate halt of
execution is in order) and others are not. If we implemented this approach we
would probably need to add this bool is_fatal as a field to struct
error_context.

Also, unless a macro is employed, the reporting of errors is not coupled with
control flow (#2) unlike promise, since PROMISE_THROW() invokes a return
statement. Of course, Peff’s code there was just a sketch instead of a
full-blown patch, but not coupling the two could easily lead to bugs in the form
of accidentally-omitted return statements, so perhaps a macro is in order if
that pattern is implemented.

The lifecycle of anything in void *data poses its own challenges. While promise
offers macros to handle the strbuf release at EOL, special handlers would need
to be written to free anything stored in void *data. Which brings me to this
feedback:

> 1) It is hard to see how we can wrap the return value of the function in a
> promise and preserve type safety. Even if we used some  kind of union the
> compiler will not warn us if the caller reads a different member to the one
> that the callee set.
>
> 2) It obscures the return type of the function and forces callers to unpack
> the value from the promise adding complexity to the calling code.

I should clarify that the promise concept is not intended to expand in scope as
far as the return value. It should remain as `int` and only ever `int`, so no
return value would be anything other than a simple number, meaning AFAIK (1)
should not apply. For (2), if auxiliary data structures (additional ints, or
structs, etc.) are outputs, they would need to be via "output parameters" passed
into the function. Those output parameters should not need to be unpacked, since
their type is preserved in the normal way.

However to return to void *data in struct error_context, the unpacking *would*
be necessary since *data doesn’t have a concrete type. Therefore, if we go with
struct error_context, the *data field should be omitted since it adds complexity
without need.

Since we are dropping async support, I am going to rename the topic from
`Promise` to `Result` as well, since the idea is equivalent to `Result<T, E>` in
Rust, aka `Either a b` in Haskell.

Regarding error messages, totally valid criticism that they are confusing to the
user as posed in the original patchset. Eventually perhaps the messages could be
reworded to increase clarity, but to avoid scope creep, let’s just not show
them. On the topic of scope, I agree with what I believe Phillip Wood implied,
that excessive refactoring should not be required to adopt this pattern. As such
I will keep this in mind in the next iteration.

There is more to discuss about strings vs error codes and the future of error
codes as well as memory allocation, but this is probably enough for now. I may
be interested to try out the error_context as well as a concept next just to see
what it would be like, but do people agree with the differential analysis so far
and changes proposed? And curious if you have any other thoughts?

- Philip

On Wed, Feb 21, 2024 at 1:03 PM Jeff King <peff@peff.net> wrote:
>
> On Tue, Feb 20, 2024 at 01:19:16PM +0100, Patrick Steinhardt wrote:
>
> > While we're already at it throwing ideas around, I also have to wonder
> > whether this would be a long-term solution towards computer-friendly
> > errors. One of the problems we quite frequently hit in Gitaly is that we
> > are forced to parse error messages in order to figure out why exactly
> > something has failed. Needless to say, this is quite fragile and also
> > feels very wrong.
> >
> > Now if we had a more structured way to pass errors around this might
> > also enable us to convey more meaning to the caller of Git commands. In
> > a hypothetical world where all errors were using an explicit error type,
> > then this error type could eventually become richer and contain more
> > information that is relevant to the calling script. And if such rich
> > error information was available, then it would not be that far fetched
> > to ask Git to emit errors in a computer-parsable format like for example
> > JSON.
>
> I think what I'm proposing (and if I understand correctly what Phillip
> was thinking) is somewhat orthogonal. I agree that structured errors are
> nice for computers to read. But it does open up a pretty big can of
> worms regarding classifying each error, especially as root causes are
> often multi-level.
>
> For example, imagine that the caller asks to resolve a ref. We might get
> a syscall error opening the loose ref. Or we might get one opening the
> packed-refs file (in a reftable world, you might imagine errors opening
> various other files). What is the structured error? Obviously it is "we
> can't resolve that ref" at some level. But the caller might want to know
> about the failed open (whether it is just ENOENT, or if we ran into
> EPERM or even EIO).
>
> Or looking at higher levels; if I ask for the merge base between A and
> B, but one of those can't be resolved, how do we communicate that error?
> It is some sort of "cannot resolve" error, but it needs to be
> parameterized to know which is which.
>
> All of those questions can be answered, of course, but now we are
> developing a micro-format that lets us describe all errors in a
> standardized way. And I think that is going to put a burden on the code
> which is generating the errors (and potentially on the consumers, too,
> if they have to decipher the structure to figure out what they want).
>
> Whereas what I was really advocating for is punting on the structured
> thing entirely. We keep our unstructured string errors for the most
> part, but we simply let the caller pass in a context that tells us what
> to do with them. That lets us keep providing specific error messages
> from low-level functions without printing to stderr or exiting, which
> higher-level code (especially lib-ified code) would not want.
>
> I think it could also be the first building block for making more
> structured errors (since those low-level callers are free to provide
> lots of details), but it doesn't have to be.
>
> -Peff