Deprecate `scipy.integrate.romberg`/`scipy.integrate.quadrature`?
Hi Team, `scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions. gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510! Thanks, Matt
+1 for deprecating them. вс, 12 нояб. 2023 г., 02:13 Matt Haberland <matt.haberland@gmail.com>:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt _______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: evgeny.burovskiy@gmail.com
+1 from me too. On Sun, 12 Nov 2023 at 14:45, Evgeni Burovski <evgeny.burovskiy@gmail.com> wrote:
+1 for deprecating them.
вс, 12 нояб. 2023 г., 02:13 Matt Haberland <matt.haberland@gmail.com>:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt _______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: evgeny.burovskiy@gmail.com
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: kaistriega@gmail.com
Hi team, Thanks for the feedback on deprecating these functions. There was support for deprecating them, but also some concerns, so the plan has changed to the following: - Deprecate `scipy.integrate.quadrature` (so we can reclaim the name for a more general-purpose function in the future). - Introduce `scipy.integrate.gauss_quad` as an alias for `scipy.integrate.quadrature`, and point users to `gauss_quad` in the `quadrature` deprecation messages. - Mark `scipy.integrate.gauss_quad` and `scipy.integrate.romberg` as legacy functions. We welcome additional feedback in gh-19510 ( https://github.com/scipy/scipy/pull/19510). Thanks, Matt On Sat, Nov 11, 2023 at 4:10 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt
On Mon, Nov 27, 2023 at 11:42 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi team,
Thanks for the feedback on deprecating these functions. There was support for deprecating them, but also some concerns, so the plan has changed to the following:
- Deprecate `scipy.integrate.quadrature` (so we can reclaim the name for a more general-purpose function in the future).
This is something we try to never do. There are very few examples of us deprecating and then removing a name, and then reintroducing it with different behavior. There has to be an exceptionally good reason to do so, and I don't see one in the PR discussion (it's quite long, so I may have missed it). I can't even think of a potential reason, because the `quadrature` name isn't special enough that you can't pick a different name for whatever you want/need here. Cheers, Ralf
- Introduce `scipy.integrate.gauss_quad` as an alias for `scipy.integrate.quadrature`, and point users to `gauss_quad` in the `quadrature` deprecation messages. - Mark `scipy.integrate.gauss_quad` and `scipy.integrate.romberg` as legacy functions.
We welcome additional feedback in gh-19510 ( https://github.com/scipy/scipy/pull/19510). Thanks, Matt
On Sat, Nov 11, 2023 at 4:10 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: ralf.gommers@gmail.com
We don't need to reclaim the name. Still, `quadrature` is a bit too generic sounding for a function that shouldn't be relied on as a general purpose routine. The best reason for wanting to reuse the name is lack of imagination. If we don't reclaim the name, what would a good name be for an improved general purpose quadrature function with several methods? I suggested the verb `quadrate`, but I don't think that's commonly used. `scipy.integrate.integrate`? On Wed, Nov 29, 2023, 1:16 AM Ralf Gommers <ralf.gommers@gmail.com> wrote:
On Mon, Nov 27, 2023 at 11:42 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi team,
Thanks for the feedback on deprecating these functions. There was support for deprecating them, but also some concerns, so the plan has changed to the following:
- Deprecate `scipy.integrate.quadrature` (so we can reclaim the name for a more general-purpose function in the future).
This is something we try to never do. There are very few examples of us deprecating and then removing a name, and then reintroducing it with different behavior. There has to be an exceptionally good reason to do so, and I don't see one in the PR discussion (it's quite long, so I may have missed it). I can't even think of a potential reason, because the `quadrature` name isn't special enough that you can't pick a different name for whatever you want/need here.
Cheers, Ralf
- Introduce `scipy.integrate.gauss_quad` as an alias for `scipy.integrate.quadrature`, and point users to `gauss_quad` in the `quadrature` deprecation messages. - Mark `scipy.integrate.gauss_quad` and `scipy.integrate.romberg` as legacy functions.
We welcome additional feedback in gh-19510 ( https://github.com/scipy/scipy/pull/19510). Thanks, Matt
On Sat, Nov 11, 2023 at 4:10 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: ralf.gommers@gmail.com
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: matt.haberland@gmail.com
But there's no rush for that. We'll open a separate issue about the new function and discuss its name when the time comes. On Wed, Nov 29, 2023, 5:14 AM Matt Haberland <matt.haberland@gmail.com> wrote:
We don't need to reclaim the name. Still, `quadrature` is a bit too generic sounding for a function that shouldn't be relied on as a general purpose routine.
The best reason for wanting to reuse the name is lack of imagination. If we don't reclaim the name, what would a good name be for an improved general purpose quadrature function with several methods? I suggested the verb `quadrate`, but I don't think that's commonly used. `scipy.integrate.integrate`?
On Wed, Nov 29, 2023, 1:16 AM Ralf Gommers <ralf.gommers@gmail.com> wrote:
On Mon, Nov 27, 2023 at 11:42 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi team,
Thanks for the feedback on deprecating these functions. There was support for deprecating them, but also some concerns, so the plan has changed to the following:
- Deprecate `scipy.integrate.quadrature` (so we can reclaim the name for a more general-purpose function in the future).
This is something we try to never do. There are very few examples of us deprecating and then removing a name, and then reintroducing it with different behavior. There has to be an exceptionally good reason to do so, and I don't see one in the PR discussion (it's quite long, so I may have missed it). I can't even think of a potential reason, because the `quadrature` name isn't special enough that you can't pick a different name for whatever you want/need here.
Cheers, Ralf
- Introduce `scipy.integrate.gauss_quad` as an alias for `scipy.integrate.quadrature`, and point users to `gauss_quad` in the `quadrature` deprecation messages. - Mark `scipy.integrate.gauss_quad` and `scipy.integrate.romberg` as legacy functions.
We welcome additional feedback in gh-19510 ( https://github.com/scipy/scipy/pull/19510). Thanks, Matt
On Sat, Nov 11, 2023 at 4:10 PM Matt Haberland <matt.haberland@gmail.com> wrote:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: ralf.gommers@gmail.com
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: matt.haberland@gmail.com
On Wed, Nov 29, 2023 at 3:05 PM Matt Haberland <matt.haberland@gmail.com> wrote:
But there's no rush for that. We'll open a separate issue about the new function and discuss its name when the time comes.
Agreed. Such a function may appear in a few months, or never - best to cross that bridge when we get to it. For now I think the only thing to decide is whether to deprecate `quadrature` or not. If it's kept for educational purposes, then all that needs doing is adding a warning to the docs about its accuracy (no new `gauss_quad` function either I'd say). And if deprecation is warranted, then just deprecate it and remove it a few releases later. Cheers, Ralf
On Wed, Nov 29, 2023, 5:14 AM Matt Haberland <matt.haberland@gmail.com> wrote:
We don't need to reclaim the name. Still, `quadrature` is a bit too generic sounding for a function that shouldn't be relied on as a general purpose routine.
The best reason for wanting to reuse the name is lack of imagination. If we don't reclaim the name, what would a good name be for an improved general purpose quadrature function with several methods? I suggested the verb `quadrate`, but I don't think that's commonly used. `scipy.integrate.integrate`?
On Wed, Nov 29, 2023, 1:16 AM Ralf Gommers <ralf.gommers@gmail.com> wrote:
On Mon, Nov 27, 2023 at 11:42 PM Matt Haberland < matt.haberland@gmail.com> wrote:
Hi team,
Thanks for the feedback on deprecating these functions. There was support for deprecating them, but also some concerns, so the plan has changed to the following:
- Deprecate `scipy.integrate.quadrature` (so we can reclaim the name for a more general-purpose function in the future).
This is something we try to never do. There are very few examples of us deprecating and then removing a name, and then reintroducing it with different behavior. There has to be an exceptionally good reason to do so, and I don't see one in the PR discussion (it's quite long, so I may have missed it). I can't even think of a potential reason, because the `quadrature` name isn't special enough that you can't pick a different name for whatever you want/need here.
Cheers, Ralf
- Introduce `scipy.integrate.gauss_quad` as an alias for `scipy.integrate.quadrature`, and point users to `gauss_quad` in the `quadrature` deprecation messages. - Mark `scipy.integrate.gauss_quad` and `scipy.integrate.romberg` as legacy functions.
We welcome additional feedback in gh-19510 ( https://github.com/scipy/scipy/pull/19510). Thanks, Matt
On Sat, Nov 11, 2023 at 4:10 PM Matt Haberland < matt.haberland@gmail.com> wrote:
Hi Team,
`scipy.integrate.romberg` and `scipy.integrate.quadrature` do not reliably hit their target accuracies, and when they do, they typically require more function evaluations and take more wall clock time than `scipy.integrate.quad` (all with default settings). This is documented in https://github.com/scipy/scipy/issues/18574#issuecomment-1703320598, which compares the accuracies, function call counts, and wall clock times on a benchmark set of ~800 integrals with known analytical solutions.
gh-19510 (https://github.com/scipy/scipy/pull/19510) proposes to deprecate these functions, suggesting `scipy.integrate.quad` as a drop-in replacement. Please let us know your thoughts either here or in gh-19510!
Thanks, Matt
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: ralf.gommers@gmail.com
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: matt.haberland@gmail.com
_______________________________________________ SciPy-Dev mailing list -- scipy-dev@python.org To unsubscribe send an email to scipy-dev-leave@python.org https://mail.python.org/mailman3/lists/scipy-dev.python.org/ Member address: ralf.gommers@gmail.com
participants (4)
-
Evgeni Burovski -
Kai Striega -
Matt Haberland -
Ralf Gommers