Skip to content

Updates master with develop before releasing v1.14.0 - #1194

Draft
Gui-FernandesBR wants to merge 144 commits into
masterfrom
develop
Draft

Gui-FernandesBR wants to merge 144 commits into
masterfrom
develop

Conversation

@Gui-FernandesBR

@Gui-FernandesBR Gui-FernandesBR commented Sep 14, 2026 •

Copy link
Copy Markdown
Member

It seems like we already have enough content to prepare a new release!
PRs #1187 and #1190 will also be included in this release.
All other PRs must wait for v1.15 or higher versions.

Issues closed by this release

Fixed on develop and labelled Awaiting release; merging this PR into master closes them.

Closes #28
Closes #340
Closes #411
Closes #437
Closes #520
Closes #524
Closes #545
Closes #604
Closes #654
Closes #767
Closes #780
Closes #953
Closes #954
Closes #963
Closes #1086
Closes #1091
Closes #1096
Closes #1106
Closes #1110
Closes #1145
Closes #1147
Closes #1151

Gui-FernandesBR and others added 30 commits July 21, 2026 21:35
CI: make changelog automation LLM-based (Gemini) and race-safe
* ENH: Resolve pressure_ISA discretization bounds TODO (#1056)

* TST: Fix test_flight and environment interpolation due to new ISA bounds

* MNT: self-document pressure_ISA bounds and strengthen discretization test

Derive the pressure_ISA discretization bounds from the standard-atmosphere
layer table (geopotential_height[0]/[-1]) instead of hardcoding -2000/80000,
and document why the grid is split around sea level. Strengthen the
discretization test with physical-sanity assertions (strictly increasing
altitude, strictly decreasing pressure, sea level sampled). No change in
numeric output.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* TST: update density doctest for new pressure_ISA discretization

The finer spline knots shift density at 1000 m by ~1e-4 (density at sea
level is unchanged since 0 m is still sampled exactly). Update the
calculate_density_profile doctest expected value accordingly.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* TST: relax noisy stream_velocity_z apogee tolerance

stream_velocity_z at apogee is a residual of the apogee-time estimation:
it is physically ~0 but swings by ~1e-4 m/s across platforms/NumPy
versions and atmosphere discretizations (e.g. -8.9e-8 on py3.10 with the
new ISA grid, -2.0e-4 with the old grid, +2.6e-4 on py3.14). The previous
atol=1e-5 was tighter than this numerical noise, making the assertion
flaky. Use atol=1e-3, which is physically negligible (vertical speed
peaks above 200 m/s) while still catching real regressions.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* STY: apply ruff markdown formatting to README code blocks

CI installs ruff unpinned; ruff 0.16 formats fenced Python blocks in
Markdown by default, so `ruff format --check .` now flags README.md on
every PR. Reformat the affected snippets (indentation, quotes, call
wrapping) to unblock the lint check. No semantic changes.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* MNT: satisfy pylint on the pressure_ISA discretization changes

CI runs pylint unpinned; three messages were attributable to this PR:
- C0415 import-outside-toplevel: move the tools import to the test module
  top level.
- R0915 too-many-statements: keep the discretization block compact so
  pressure_ISA stays within the 25-statement limit.
- C0302 too-many-lines: the terser block keeps environment.py under the
  3050-line module limit.

np.append over two linspaces yields the same grid as the previous
np.concatenate, so all numeric results are unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* ENH: Add native Meteomatics API support to the Environment class

Adds a new "meteomatics" atmospheric model to Environment.set_atmospheric_model,
porting and generalizing the implementation from the EuRoC-Dev repository.

- fetchers.py: fetch_meteomatics_token + fetch_atmospheric_data_from_meteomatics
  authenticate with username/password (short-lived token), query temperature,
  pressure and wind components by height above ground level, grouping the
  parameters to respect the account's per-request limit.
- environment.py: process_meteomatics_atmosphere converts the height-AGL data to
  above-sea-level profiles using the Environment elevation; set_atmospheric_model
  gains username/password kwargs (falling back to METEOMATICS_USERNAME /
  METEOMATICS_PASSWORD env vars); save/load handles the new model type.
- Network requests use timeouts, do not retry deterministic 4xx failures, and
  surface actionable RuntimeError messages.
- Tests fully mock the API (no real requests, no charges); docs and changelog
  updated.

Closes #545

* MNT: simplify Meteomatics fetcher and profile assembly

Follow-up cleanups from the code review of the Meteomatics support:

- Collapse the duplicated request/error ladder of the login and data
  endpoints into a single `_meteomatics_request_json` helper.
- Have `_build_meteomatics_parameters` return a {parameter: (profile,
  height)} mapping so the response parser no longer regex-parses back
  the strings this module just built. Drops the regex constant and the
  variable-to-profile table.
- Generalize `to_profile_array` to several profiles at once, so the
  wind u/v grid intersection reuses it instead of repeating it inline.
- Reuse the existing `__validate_datetime` helper for the launch-date
  check, and simplify the model default to `model or "mix"`.
- Normalize `atmospheric_model_type` once in `from_dict`. The type is
  stored as the user spelled it, so the previously case-sensitive
  `match` and `== "ensemble"` branches silently dropped the ensemble
  arrays for an Environment built with `type="Ensemble"`.
- Trim the elevation warning and raise it before the API call, so it is
  shown even when the request later fails.

* MNT: address Meteomatics review comments

Two points raised in the PR review:

- `fetch_atmospheric_data_from_meteomatics` stamped the instant with a
  trailing "Z" without normalizing the timezone, so an aware datetime in
  a non-UTC zone was sent as the wrong instant. Aware datetimes are now
  converted to UTC; naive ones are documented as assumed UTC.
- Degenerate sampling arguments (`query_limit=0`, resolutions below 2)
  reached `range()`/`linspace()` and failed with an opaque low-level
  error after the login had already been paid for. They are now
  validated up front by `_validate_meteomatics_sampling`.
- `process_meteomatics_atmosphere` silently coerced any non-string model
  to "mix", hiding a mistake such as passing a Dataset or a path as
  `file` and querying the wrong model. It now accepts None (default) or
  a string, and raises otherwise.


* ENH: Refactor Meteomatics integration into MeteomaticsFetcher class and clean up environment method

* ENH: Decompose fetchers.py into rocketpy/environment/fetchers package with dedicated submodules
Both examples build a generator with np.random.default_rng(seed) and drop it,
then sample from the process-global np.random. reset_seed is a no-op, so a
sampler written by following this page ignores random_seed. The study runs and
the numbers look reasonable; only a second run with the same seed shows they
were never reproducible.

The bivariate generator needs two more things. It caches 1000 samples, so
reset_seed has to discard them or the first 1000 draws after a reseed still
come from the generator that was replaced. And its refill test compares
samples_generated against used_samples, which only becomes true after the
cache has run out, so a study longer than the cache silently got short lists:
sample(n_samples=1) returns [] at simulation 1001 and dict_generator raises
IndexError on [0].

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
…1078) (#1084)

Isolates the three VTK/PyVista off-screen animation tests into their own workflow step so a native crash in one of them no longer kills the whole integration run, adds a bounded retry for the crash, forces Mesa software rendering on Linux, and sets fail-fast: false so one platform no longer cancels the rest of the matrix.

Closes #1078
…411) (#1085)

A motor whose thrust curve starts at t > 0 never left the rail: the rail phase's only time nodes were [t=0, max_time], so LSODA (max_step defaults to inf) took one huge step over the entire burn and the rocket never accelerated. __setup_phase_time_nodes now forces solver stops at ignition and burn-out when burn_start_time > 0, leaving ordinary motors byte-for-byte unchanged.

Also moves Codecov's coverage config out of .github/workflows/codecov.yml (where Actions was trying to execute it as a workflow, failing on every run) to .codecov.yml.

Closes #411
StochasticRocket.create_object built a Parachute from the full sampled draw, then discarded it and built a second one from six of its ten fields. radius, height, porosity and drag_coefficient never reached the rocket: the second Parachute re-derived radius from cd_s and the default drag coefficient, and height fell back to that radius. Attaching the sampled object itself fixes it and stops Parachute.__init__ running twice per parachute per simulation.

Breaking: radius and height feed the parachute added-mass term in flight.py, so descent dynamics and landing points change for any study whose parachute carries geometry.

Closes #1094
#1100)

* MNT: do not let one Python version cancel the other in the slow matrix

#1084 added fail-fast: false to the main test matrix while this was open, so
the only half left is the slow one. Same reason: 3.10 failing says nothing
about 3.14, so cancelling it costs a result and saves nothing worth having.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

* MNT: retry the VTK tests on SIGSEGV, not only on macOS SIGBUS

#1084 retries the animation tests when they die on 138, which is SIGBUS on
macOS. Counting the last 40 Tests runs, the crash was 139 eight times and 138
twice, so the common case fell straight through the retry. Linux SIGBUS is 135
rather than 138, so that missed as well.

Also sets fail-fast: false on the slow matrix, which #1084 left out. 3.10
failing says nothing about 3.14, so cancelling it costs a result and saves
nothing worth having.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

---------

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
Co-authored-by: Gui-FernandesBR <63590233+Gui-FernandesBR@users.noreply.github.com>
* ENH: Add Qodo PR-Agent workflow using Google Gemini

* FIX: point PR Agent workflow at the renamed action repo

Codium-ai/pr-agent-action no longer exists (the project moved to
the-pr-agent/pr-agent), which made the "Run PR Agent" check fail with
"Unable to resolve action". Updates the action reference, grants the
contents: write permission it now requires, and sets the Gemini key
via the GOOGLE_AI_STUDIO.GEMINI_API_KEY env var per the new action's docs.
* CI: build the docs for pull requests into develop as well

`branches` filters on a pull request's base, and almost every PR here is opened
against develop: 25 of the last 30, with 3 against master. So the build that
runs with -W --keep-going saw a docs change for the first time in a release
batch, well away from whatever caused it.

The path filter already keeps this off PRs that cannot affect the docs, and
`rocketpy/**` is in it because docstrings feed the autodoc reference, so a
docstring edit merged into develop could break the API pages unnoticed too.

`push` is left on master alone. The pull request check is where the feedback is
worth having, and running both would double the cost for a second opinion on
the same commit.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

* Apply suggestion from @Gui-FernandesBR

---------

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
Co-authored-by: Gui-FernandesBR <63590233+Gui-FernandesBR@users.noreply.github.com>
Every sampler on a model was reset with the model's seed, so two backed by
default_rng started from identical state and drew identical underlying values.
Not nearly identical, the same to every digit: two Gaussians with different
means and spreads both produced the deviate 0.466220770577340.

A study varying two parameters that way is varying one, and the correlation it
reports between them is an artefact of the seeding.

Each sampler now gets a child derived from the model's seed and the input's
name. Keyed by name rather than position so declaring another parameter does
not move the streams of the ones already there, and crc32 rather than hash
because hash is not stable across processes.

The documented wind X/Y wrappers are unaffected. Their correlation comes from
sharing one samples_list, not from sharing a seed, so handing them separate
children leaves it intact: measured 0.7010 against the covariance's 0.6981.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
Two problems with the first version of this, both found in review.

CRC32 is 32 bits, and a collision puts two samplers back on one stream, which
is the bug the keying exists to prevent. `wd4s4xka50` and `p56cjcee10` are both
valid identifiers with CRC32 1560575156, and both derived the same seed. The
name is length-prefixed into spawn-key words now, which no two names share, and
the child is kept at its full 128 bits to match the Monte Carlo seeding rather
than being cut to 64.

Samplers can also share one generator on purpose, as the documented wind pair
does, and each reset overwrites the last. With one seed per name, whichever was
reset last decided the stream, so the same seed meant different runs depending
on the order the model was declared in. Seeding is its own pass over sorted
names now.

The pass is separate from the validation loop deliberately. That loop's order
sets __dict__, and so the order every other input is drawn in, so sorting it
would have moved the samples of every model with a tuple in it.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
Two wrappers can share one generator on purpose, as the documented wind pair
do. One seed per name reset that generator once per wrapper, so every seed but
the last was discarded and the group's stream was decided by whichever member
sorted last. Adding a third wrapper to the same generator therefore moved the
first two, which name keying is meant to prevent.

CustomSampler gains a `seed_group` property, `self` by default, so a wrapper
can say which generator it shares. Members of a group are seeded once between
them, with the seed derived from all their names rather than from whichever
went last. The documented wind wrappers declare it.

Before, resetting six times for three wrappers and moving the pair when a third
arrived. After, once, and adding an independent sampler leaves the group where
it was.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
Only RuntimeError was caught, and the seed handed over is now 128 bits, which
the legacy numpy.random.RandomState refuses:

    ValueError: Seed must be between 0 and 2**32 - 1

Before this branch a sampler received the model's seed, usually a small int,
so RandomState took it. A sampler built on RandomState therefore breaks here,
and used to break with a bare ValueError that named nothing.

The seed stays 128 bits, since that is what keeps the streams apart and what
default_rng, the documented choice, takes. The error now says which input the
sampler belongs to and keeps the original as its cause.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
…discards

Three things from review, all small.

The documented bivariate generator filled a 1000-pair cache inside reset_seed.
With per-index seeding that reset happens once per simulation, so a study built
on this example generated a thousand pairs and used one, every time. 0.099 ms
each, about 10 s over 100k simulations. `top_up` already fills the shortfall on
first use, so the eager fill is gone and the cache starts empty.

The group reset went through the first member rather than the group, which
assumes every member resets identically and keeps nothing of its own. The group
holds the shared state, so it is reset directly when it knows how, and the
member is the fallback.

`_sampler_seed` now sorts the names itself. The caller does today, and a future
one that forgets would hand a single group two different seeds.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
Two rules the property invites breaking, both silent.

Identity has to be stable. Building the answer on each call, which returning
from a property makes easy, gives every member a different identity and puts
each back in a group of its own: a two-member group goes from one reset to two.

A group belongs to one model. Declaring the same generator on two models has
them both seed it, and the later one wins, which is the overwrite the grouping
exists to prevent.

The documented wind pair already returns a stored attribute, so the example
teaches the stable form.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
The automation that normally writes it cannot run on a pull request from a
fork, which is #1101, so this one is by hand.

It is a breaking change and the entry says so: fixed-seed CustomSampler
baselines move, and a sampler built on the legacy RandomState has to move to
default_rng because the seed it now receives is 128 bits wide.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
`scipy>=1.0` is not true. `monte_carlo.py` imports `scipy.stats.bootstrap` at
module level, and the SciPy 1.7.0 release notes are where that arrives, so
anything from 1.0 to 1.6 satisfies the floor and then fails on `import
rocketpy`.

`numpy>=1.13` is not true either, and the binding constraint turned out to be a
sibling rather than NumPy itself: `matplotlib>=3.9.0` requires `numpy>=1.23`.
SciPy then has to be new enough to allow that, and 1.7.2 caps NumPy at
`<1.23.0`, so 1.8 is the first that composes.

Verified rather than reasoned. On Python 3.10 with numpy 1.23.0 and scipy
1.8.0 pinned and everything else current, `import rocketpy` works and
`tests/unit/simulation` with `tests/unit/stochastic` is 162 passed, 5 skipped.
At numpy 1.21.3, which is where NumPy's own cp310 wheels start, matplotlib
refuses to import.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
…1103)

* BUG: accept the callable parachute triggers the docstring promises

The `or` sat inside the isinstance call rather than beside it:

    isinstance(member, (str, int, float) or callable(member))

A non-empty type tuple is truthy, so the expression short-circuited to the
tuple and callable(member) was never evaluated. The check reduced to
isinstance(member, (str, int, float)), and a callable is none of those.

Parachute takes a callable trigger and Flight calls it, and the docstring three
lines above says "a list of callables, string 'apogee' or ints/floats". Only
the stochastic wrapper refused one. The two non-callable forms passed
throughout, which is why the tests never caught it.

Left as an assert to match the other fourteen in these two modules, and because
raising a different type would break anyone catching AssertionError.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

* BUG: refuse the trigger forms Parachute cannot use

Moving the `or` let callables through, but left four things the check should
never have accepted.

`["banana"]` passed here and then `Parachute` raised ValueError, so the wrapper
only moved the failure to create time. `[True]` is worse: bool is an int, so it
went through as a height of one metre. `[]` passed because `all([])` is True.
And `numbers.Real` replaces `(int, float)`, which took numpy.float64 because it
subclasses float and refused numpy.int64 because it subclasses neither.

The check is raised rather than asserted. `python -O` strips an assert outright,
and this is the only thing between those triggers and a Parachute that either
refuses them later or misreads them. Still an AssertionError, so nothing that
catches it has to change.

The docstring claimed a tuple form that was never implemented; it now describes
what the code does.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

* BUG: do not accept numpy integers the Parachute will refuse

Widening the height check to numbers.Real was the same mistake as accepting
"banana": Parachute checks isinstance(trigger, (int, float)), so numpy.float64
passes because it subclasses float and numpy.int64 raises ValueError because it
subclasses neither. Letting them through here only moved the failure to
create_object.

Back to (int, float), matching that check rather than improving on it, and the
test that asserted numpy.int64 was accepted now asserts both ends refuse it.

The asymmetry is Parachute's rather than this wrapper's and is worth fixing
there, where widening the check would not strand anything downstream.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

* TST: use the state length Flight actually passes

The trigger tests built a 14-element state. Flight passes 13:
x y z vx vy vz e0 e1 e2 e3 wx wy wz. Only y[5] is read, so both worked, but
the test is there to document the contract and was documenting it wrongly.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

* DOC: add the changelog entry for this branch

By hand, because the automation cannot run on a pull request from a fork
(#1101).

No documented API breaks, but the observable behaviour does: an invalid string,
an empty list or a boolean trigger now fails during StochasticParachute
validation rather than later in Parachute construction, or silently becoming a
one-metre height trigger.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

---------

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
BUG: give each CustomSampler its own stream instead of the model's seed

Every CustomSampler on a model was reset with the model's own seed, so two
samplers backed by default_rng started from the same state and drew the same
underlying deviate. Each input now gets a stream keyed by its name, and
samplers that share one generator declare a seed_group so the group is seeded
once between them.

Merged manually rather than through the button: the CHANGELOG conflicted with
#1103, and the resolution keeps both entries.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* ENH: Add opening shock force estimation to Parachute class (#1050)

Adds an opening_shock_coefficient parameter and a
calculate_opening_shock_force method to estimate the peak transient
force during parachute inflation, following the simplified model in
Knacke's Parachute Recovery Systems Design Manual (1992, Section 5.5).

Closes #161

Co-authored-by: ArthurJWH <167456467+ArthurJWH@users.noreply.github.com>

* ENH: Moving opening shock force function to utilities

* DOC: Updated the CHANGELOG

* DOC: Removed cross-reference from previous Parachute method

---------

Co-authored-by: Matanski <Matangrod@outlook.co.il>
#1104 and #1108 both came from a fork, so `Populate Changelog` failed on
`RELEASE_TOKEN` being withheld and neither got an entry. #1101 tracks the
workflow bug; these two lines are the entries it would have written.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Gui's point on the review. `python -O` strips an assert, and this is what keeps
a non-sampler out of the model, so it has to be a raise. Same shape as #1103,
which took the identical route for the parachute triggers.

AssertionError is kept rather than swapped for TypeError, because the docstring
on develop already documents it and a caller catching it should keep working.

Two tests. One is the behaviour; the other runs a child interpreter under -O,
since that is the mechanism and the plain test passes either way.

Note this module carries thirteen more asserts on develop, none of them mine.
Happy to send them separately if you want the same treatment there.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
`stochastic_model.py` has carried this as a TODO:

    # TODO: Stop using assert in production code. Use exceptions instead.

The reason it matters is that the optimiser removes them, so the checks stop
running and malformed input reaches the model:

    StochasticModel(obj, mass=("not a number", 0.5))

    python      AssertionError
    python -O   accepted, mass = ('not a number', 0.5, <Generator.normal>)

That model then fails somewhere later with something that does not point back
at the tuple that caused it.

Twenty one of them, all in stochastic/: thirteen in stochastic_model.py, three
each in stochastic_environment.py and stochastic_flight.py, one in
stochastic_aero_surfaces.py, and the one in stochastic_parachute.py that #1103
did not reach. Converted mechanically through the AST, so the condition and the
message are the ones that were there.

AssertionError is kept rather than swapped for TypeError or ValueError. The
docstrings document it, and callers catching it should keep working. Changing
the type is a separate decision from making the check run at all.

Two kinds of test. One reads each module and fails on a reintroduced `assert`,
which is the mechanism. The other runs a child interpreter under -O, because in
process the assert is still compiled in and the check would pass either way.

Also moves two imports in test_custom_sampler.py to the top of the file, which
pylint flags as C0415 on develop today.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
MateusStano and others added 2 commits November 16, 2026 21:04
#1195)

* BUG: use the angular rates in the solid propulsion equations of motion

Flight.u_dot zeroed omega1, omega2 and omega3 right after unpacking the
state, so the aerodynamics and the Euler equations saw a rocket that
never rotates: no aerodynamic damping at all. A pitch rate produced no
restoring moment and a canted fin spun the rocket up without bound
(1000 rad/s for a 1.5 degree cant on Calisto). Every rocket with a
SolidMotor is mapped to these equations, so every such 6-DOF flight was
affected. Introduced with the 3-DOF work in #745.

With the rates back, the solid propulsion equations agree with the
generalized ones to 0.25% on apogee and 0.2% on roll rate.

Tests: each body axis rate must produce a damping angular acceleration
with both sets of equations, the two must agree when rotating, and a
rolling flight's roll rate must stay bounded and track the airspeed.

* BUG: fix the propellant offset axis and the inertia rate in Flight.u_dot

Two old errors in the solid propulsion equations that act once the
angular rates are real:

- The propellant offset terms of the linear acceleration were written for
  an offset along body axis 1. The propellant sits on the rocket axis
  (axis 3), so a roll rate pushed the rocket sideways during the burn
  (about 30 m/s2 at 20 rad/s on Calisto).
- The rate of change of the reduced-mass inertia used
  (rocket_dry_mass - 1) where the derivative of mu * b**2 gives
  rocket_dry_mass**2.

Two burn-phase tests compare the equations with u_dot_generalized.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
* STY: put the pylint disable in tools.py on its own line

The line was too long for ruff format --check.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* BUG: fix the tip-off phase and run it without a flag

The phase from #920 now runs in every 6 DOF flight with the standard
equations of motion whose two rail buttons start on the rail. The
use_udot_rail2 flag is removed.

- udot_rail2 is rewritten: same model, no closure, no numpy solve and no
  cache written by u_dot_generalized on every call.
- Rail exits are found with brentq on the solver's dense output instead
  of one cubic fitted over the last step, which misplaced the exit on
  long steps and could raise "No valid roots found".
- The accelerations reported inside the window are the constrained ones.
- The lower rail button forces cover the window and are saved with the
  flight.
- between_rails_* is renamed to tip_off_end_*, and tip_off_end_time is
  saved with the flight.
- A parachute opening inside the window is no longer handed back to free
  flight, tip_off_duration is never negative, and a rail shorter than the
  nozzle to lower button distance skips the phase.
- The on-rail button forces select their points by time, since flights
  with controllers store no post-processed point at event times.

The tests compare the phase with an independent rigid-body model of a
button sliding on the rail (pitch angle, pitch rate and button force).
Pinned values that depend on the trajectory are re-baselined.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* DOC: update the tip-off page for the always-on phase

Removes the flag, corrects the sign of the r_CM x omega_dot terms after
#1196, states the limits of the model (no friction, no contact between
the rocket body and the rail), describes the outputs and adds a
verification section with the reference of Peng et al. (2017).

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
Gui-FernandesBR and others added 15 commits November 17, 2026 14:10
Addresses the Monte Carlo fixes in #1209. Simplifies study seeds and checkpoint recovery, preserves stochastic configurations, and adds regression coverage and migration notes.
Keep the base flight's terminate_on_apogee and use_udot_rail2 in StochasticFlight and MonteCarlo, accept NumPy scalars in initial_solution, and stop a checkpointed log's length from hiding a stray index. Seeds a study derives for itself no longer count as object seeds, so a second study does not warn about them. Drop docs/TODO.md and fold the changelog's breaking-changes section into Changed. #1209

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW574nZTvpQsbhf7kAaFz
MonteCarlo wrote out the Flight constructor a second time, so every new Flight argument had to be added in two places. It now passes its sampled rocket and environment to create_object. #1209

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Output rows now record the study seed, and an append checks that both logs carry the same seed row by row. Before, inputs from one study and outputs from another resumed as one. Collector names reserved for row metadata come from the same constant that keeps them out of results. #1209

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Review of #1213. Resuming guessed that a legacy log whose indices all start at 1 was one-based, but an interrupted v1.13 parallel log also holes index 0 and was then extended past the target instead of filled. The study now plans every index below the target that the checkpoint does not hold, hands that explicit list to the progress monitor, and checks that the logs hold exactly the checkpoint plus the plan.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Review of #1213. Each simulate_convergence batch appended through a full parse of both logs, so a study re-read its whole history once per batch. A run now remembers the size and modification time of the logs it left behind with their seed and indices, and the next append reuses them while the files are unchanged; any other change reads the logs again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
…ller

Review of #1213. Only the serial path cleared the seeds a caller had set on the stochastic objects, because parallel workers reseed copies, so a study run in parallel warned again on the next run. The study now clears them in the parent before either path runs, and the object-seed and legacy-log warnings point at the first frame outside rocketpy, which is the user's line also when they come through simulate_convergence.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Review of #1213. Every sampler seed was cut to 32 bits so RandomState could take it, which makes two of a study's sampler streams collide after about 2**16 draws. Samplers now get the full seed and are retried with its low 32 bits only when they raise ValueError, as RandomState does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Review of #1213. StochasticFlight told a launch from an explicit state apart by comparing initial_derivative with the bound method udot_rail1, which breaks with any wrapper or rename of that method. Flight now records _initial_solution_given before it fills initial_solution in, and StochasticFlight reads it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Review of #1213. A base flight built with terminate_on_apogee=True silently stopped every sampled flight at apogee, which drops descent, drift and impact outputs from the study. StochasticFlight still keeps the value, but now warns and names terminate_on_apogee=False as the way out; a value the caller passes does not warn.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
Review of #1213. The lint job failed on four points the review fixes introduced: an unused Flight import left once flights were built through create_object, StochasticFlight.__init__ over the statement limit after the terminate_on_apogee warning, a test class without a docstring and an empty-list comparison. The warning moves into _resolve_terminate_on_apogee and its test now checks that it still points at the caller.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW576acgQ8pLTgu6VDQR2
The tip-off phase now always runs and Flight no longer has the flag, so
every StochasticFlight.create_object failed on develop with an
AttributeError. Related to #1209.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
The generated seed was 128 bits and is written on every row, so
pandas.read_json(lines=True) failed with "Value is too big!" on files
v1.13 wrote fine. A seed passed by the caller is still taken as given.
Related to #1209.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
MateusStano and others added 8 commits November 19, 2026 18:26
…ins (#1218)

* ENH: base the tube fins on ring wing measurements and drop add_tube_fins

Follow-up of #1144 for the v1.14.0 release.

- Normal force: Ribner's result that a ring wing lifts twice as much as an
  elliptic flat wing spanning its diameter, with Helmbold's equation for
  that wing. Within 3.1% of the five lift slopes measured in NACA TN 4117.
  The lifting-line formula used before is 11 to 68% higher on those points.
- Center of pressure: the positions measured in NACA TN 4117, interpolated
  on inner diameter over length, in place of a fixed quarter of the length.
- The 20 degree cap on the normal force is removed. No other surface has a
  stall model.
- Roll damping: each tube takes the normal force its sideways motion
  produces.
- Rocket.add_tube_fins and Rocket.tube_fins are removed. Tube fins are
  added with Rocket.add_surfaces.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* DOC: describe the tube fin model, its data and its limits

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* DOC: say where the Prandtl factor is applied in AeroSurface._beta

Only the fins use it. The docstring also named the nose cone and the
tails, which stopped using it in 97ea9ae.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* DOC: link the tube fin references

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
…out extras

Parsing the EuRoC reanalysis takes most of their ~200 s, and cutting the
files to the fixture's window only saved ~25%, so they go back under
`slow`. Without windrose, ipywidgets, IPython or jsonpickle they now skip
instead of failing, and the fixture pins Europe/Lisbon so test_values no
longer depends on timezonefinder. Related to #1209 and #1176.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
Only the apogee is compared with the measured flight. The max speed and
impact point are RocketPy's own output, so they are now named as a
regression check, and the max acceleration check is dropped because it
pinned a 1 us solver spike at burnout. Related to #1209.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
The maps extra in pyproject.toml already has it, but make install and the
development setup guide install from this file, so plot_trajectory_map
did not work in that environment. Related to #1209.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
The changelog workflow runs only after a merge, so a broken
update_changelog.py was found only on develop. Its test file needs no
network or extra packages. Related to #1209.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
TST: finish the v1.14.0 test checklist item
Rocket.static_margin became a read-only property, so assigning it raised AttributeError. The setter warns and keeps the value until the next geometry, mass or surface change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XDW575d9C9zpS7Z1v2KXVq
BUG: accept static_margin assignment with a DeprecationWarning
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment