Following up on discussions from #15, this is a first attempt at implementing a Kalman Filter/Smoother using Stone Soup.

This PR aims to contribute the following:

• A TrajectorySmoother base class, which (as advised) is conceptually similar to TrajectoryGeneralisers;
• A KalmanSmootherCV smoother, which employs a linear Kalman Filter/Smoother with the assumption of a Constant Velocity model (other models also exist in Stone Soup, but for now the CV model has been used as a starting point);
• A tutorial notebook which showcases how the above class can be used

This PR is work in progress and is open to modifications based on the thoughts and suggestions of the maintainer(s). Its main purpose is to demonstrate basic capabilities and motivate discussions.

I'm trying to use moving pandas to process GPS data stored on a PostGIS database. I'm connecting to PostGIS with psycopg2 I load a table to a geodataframe with geopandas. I loaded the geodataframe to trajectories and if I print the collection I get TrajectoryCollection with 34 trajectories so it seems to be ok I get the first trajectory with my_traj = traj_collection.trajectories[1]

Problem 1) If I print (my_traj) I get an error AttributeError: 'DataFrameGroupBy' object has no attribute 'geometry'

On PostGIS my geometry column is named geom, I tried to rename the geom to geometry on the geodataframe prior to creating the TrajectoryCollection, It seems to work but then I get two geomemetry fields, geom and geometry

If I print the trajectory again, I get this error. AttributeError: 'numpy.int64' object has no attribute 'to_pydatetime'

Problem 2) If I try to store speed with traj_collection.add_speed(overwrite=True) I get an error message, the last line is AttributeError: 'float' object has no attribute 'total_seconds'

Could you help me to solve this problem? Thanks!

Hi,

I have been following the tutorials with no issues, however, after finishing them all, I have tried to use hvplot to plot the TrajectoryCollection made from my dataframe, however I found that it was not returning any sort of graph: Using len(traj_collection) shows it has a length of 0, so no trajectories are being made for some reason but I am not sure why, I have followed exactly as is shown in the tutorials?

Thanks, Jacob

I am trying to add the clip approach in the 2-ship-data.ipynb notebook by adding below the get_intersecting chunk the following:

all_trajectories = traj_collection.trajectories
arrivals = [ traj.clip(area_of_interest) for traj in all_trajectories ] 

And I am getting the following error & I cannot understand why ?

 15        if type(row['spatial_intersection']) == LineString:
 16         -> pt0 = Point(row['spatial_intersection'].coords[0])
 17             ptn = Point(row['spatial_intersection'].coords[-1])
 18             t = row['prev_t']

IndexError: list index out of range

Closes #78

Finalise the work on separating the plotting libraries to make them optional. This includes:

1. Remove references to cartopy which isn't used
2. Update setup.py to parse requirements.txt files
3. Have a base requirements.txt, a via-requirements.txt and a dev-requirements.txt which install
   1. The base libraries required for movingpandas (requirements.txt)
   2. The additional visualisation libraries for the hvplot function (viz-requirements.txt)
   3. The libraries required for development, specifically unit testing libraries (dev-requirements.txt)
4. Include optional installs in setup.py so you can:
   • pip install movingpandas
   • pip install "movingpandas[viz]"
   • pip install "movingpandas[dev]"
5. Update the unit tests so that testing the hvplot function is skipped when holoviews is not installed
6. Create a tox test script to test Python 3.7 and 3.8 with and without the visualisation libraries
   • This uses the tox-conda library to use conda for installs rather than the standard venv
7. Update the travis configuration (including tox-travis) to use tox for testing and to combine the coverage report
   • Unit tests should be run with tox from the command line, and you should end up with 6 skipped tests - 3 each from the 3.7 and 3.8 environments. Be warned - this can take a while to set up the environmnents the first time around! And travis will take 10 - 12 minutes to run!

Hi, this is a part of https://github.com/pyOpenSci/software-review/issues/18.

Naive installation As a naive user, if I make a new python environment and do pip install movingpandas, the installation fails on Mac.

Collecting movingpandas
  Downloading movingpandas-0.2rc1.tar.gz (20 kB)
Collecting numpy
  Downloading numpy-1.18.1-cp38-cp38-macosx_10_9_x86_64.whl (15.2 MB)
     |████████████████████████████████| 15.2 MB 5.2 MB/s 
Collecting matplotlib
  Downloading matplotlib-3.1.2-cp38-cp38-macosx_10_9_x86_64.whl (13.2 MB)
     |████████████████████████████████| 13.2 MB 4.1 MB/s 
Collecting shapely
  Using cached Shapely-1.6.4.post2.tar.gz (225 kB)
Collecting pandas
  Using cached pandas-0.25.3-cp38-cp38-macosx_10_9_x86_64.whl (10.3 MB)
Collecting geopandas
  Using cached geopandas-0.6.2-py2.py3-none-any.whl (919 kB)
Collecting hvplot
  Downloading hvplot-0.5.2-py2.py3-none-any.whl (2.5 MB)
     |████████████████████████████████| 2.5 MB 4.5 MB/s 
Collecting cartopy
  Using cached Cartopy-0.17.0.tar.gz (8.9 MB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... error
  ERROR: Command errored out with exit status 1:
   command: /Users/martin/anaconda3/envs/movingpandas/bin/python /Users/martin/anaconda3/envs/movingpandas/lib/python3.8/site-packages/pip/_vendor/pep517/_in_process.py get_requires_for_build_wheel /var/folders/d0/qyf0yvzd0n9ctn2wr3r4l1vw0000gn/T/tmpcf10m5i7
       cwd: /private/var/folders/d0/qyf0yvzd0n9ctn2wr3r4l1vw0000gn/T/pip-install-9up68k4a/cartopy
  Complete output (1 lines):
  Proj 4.9.0 must be installed in Conda environment "movingpandas".
  ----------------------------------------
ERROR: Command errored out with exit status 1: /Users/martin/anaconda3/envs/movingpandas/bin/python /Users/martin/anaconda3/envs/movingpandas/lib/python3.8/site-packages/pip/_vendor/pep517/_in_process.py get_requires_for_build_wheel /var/folders/d0/qyf0yvzd0n9ctn2wr3r4l1vw0000gn/T/tmpcf10m5i7 Check the logs for full command output.

Using conda install -c anitagraser movingpandas works, but you'll get a mixture of channels (conda-forge and defaults), which is known to cause issues. Moreover, you get geopandas from defaults in older version that is on conda-forge.

It is worth documenting possible issues and workarounds (strict channel priority) as the default installation instructions do not follow GeoPandas recommendations.

Dev installation

Current dev installation instructions expect using Anaconda. I'd say that most dev people will just make a new environment using conda env create -f environment.yml, but it has to be derived from the Aanconda instructions. I'd just add a sentence saying that dev env should use environment.yml and keep Anaconda instructions as one of the options only.

Minor question - conda installs movingpandas 0.1.dev2, while the review issue says Version submitted: 0.2. Should I assume that 0.2 will be released after the review?

Sad error here. Clean install into dedicated environment on Windows 10. environment.yml records version as

movingpandas=0.1.dev2=py37h39e3cac_0

When trying to use the TrajectoryManager class from the tutorial it fails saying the module has no such attribute.

I can't see any releases on github so wondered if conda has lagged behind github?

Currently the pip installation is raising the follow error:

21:09 $ pip install movingpandas
ERROR: Could not find a version that satisfies the requirement movingpandas (from versions: none)
ERROR: No matching distribution found for movingpandas

Could apply tools such as black and flake8 on the repo. This can then be added to as a hooks before a commit is made e.g. https://github.com/geopandas/geopandas/blob/master/.pre-commit-config.yaml. Can also have checks on PRs.

More info: https://docs.geopandas.org/en/latest/community/contributing.html#style-guide-linting

The application for which I'm using movingpandas sometimes requires me to handle Trajectorys having a single point, but movingpandas requires each Trajectory to have at least two points.

What are your thoughts about this? I'm happy to submit a PR if you are amenable to the idea.

I think the following plan could work:

• Define a TrajectoryCollection.allow_single_point_trajectory flag that defaults to False
• Define a Trajectory.allow_single_point flag that defaults to False
• Update relevant methods to handle single point Trajectorys

I think the biggest issues would be places where multiple points are assumed, such as calculating speed. I'm not sure what the best behavior in that situation would be (None, 0, infinite, raise error, something else?).

What are your thoughts/concerns? If I'm alone on this, I'm happy to just inherit from the Trajectory and TrajectoryCollection classes inside my application.

Thank you!

Hi,

Unlike other generalization functions, TopDownTimeRatioGeneralizer returns an error message : AttributeError: The 'Series' object does not have a 'geometry' attribute. No matter that I use a trajectory or a trajectorycollection.

topDownGener=mpd.TopDownTimeRatioGeneralizer(collection).generalize(tolerance=0.0001)

`--------------------------------------------------------------------------- AttributeError Traceback (most recent call last) ~\AppData\Local\Temp/ipykernel_39264/2560778543.py in ----> 1 topDownGener=mpd.TopDownTimeRatioGeneralizer(collection).generalize(tolerance=0.0001)

~\anaconda3\envs\MovingPandas\lib\site-packages\movingpandas\trajectory_generalizer.py in generalize(self, tolerance) 42 return self._generalize_traj(self.traj, tolerance) 43 elif isinstance(self.traj, TrajectoryCollection): ---> 44 return self._generalize_traj_collection(tolerance) 45 else: 46 raise TypeError

~\anaconda3\envs\MovingPandas\lib\site-packages\movingpandas\trajectory_generalizer.py in _generalize_traj_collection(self, tolerance) 49 generalized = [] 50 for traj in self.traj: ---> 51 generalized.append(self._generalize_traj(traj, tolerance)) 52 result = copy(self.traj) 53 result.trajectories = generalized

~\anaconda3\envs\MovingPandas\lib\site-packages\movingpandas\trajectory_generalizer.py in _generalize_traj(self, traj, tolerance) 239 240 def _generalize_traj(self, traj, tolerance): --> 241 return Trajectory(self.td_tr(traj.df.copy(), tolerance), traj.id) 242 243 def td_tr(self, df, tolerance):

~\anaconda3\envs\MovingPandas\lib\site-packages\movingpandas\trajectory_generalizer.py in td_tr(self, df, tolerance) 252 dy = df.geometry.iloc[-1].y - df.geometry.iloc[0].y 253 --> 254 dists = df.apply( 255 lambda rec: self._dist_from_calced( 256 rec, df.index.min().to_pydatetime(), df.geometry.iloc[0], de, dx, dy

~\anaconda3\envs\MovingPandas\lib\site-packages\geopandas\geodataframe.py in apply(self, func, axis, raw, result_type, args, **kwargs) 1388 @doc(pd.DataFrame) 1389 def apply(self, func, axis=0, raw=False, result_type=None, args=(), **kwargs): -> 1390 result = super().apply( 1391 func, axis=axis, raw=raw, result_type=result_type, args=args, **kwargs 1392 )

~\anaconda3\envs\MovingPandas\lib\site-packages\pandas\core\frame.py in apply(self, func, axis, raw, result_type, args, **kwargs) 8831 kwargs=kwargs, 8832 ) -> 8833 return op.apply().finalize(self, method="apply") 8834 8835 def applymap(

~\anaconda3\envs\MovingPandas\lib\site-packages\pandas\core\apply.py in apply(self) 725 return self.apply_raw() 726 --> 727 return self.apply_standard() 728 729 def agg(self):

~\anaconda3\envs\MovingPandas\lib\site-packages\pandas\core\apply.py in apply_standard(self) 849 850 def apply_standard(self): --> 851 results, res_index = self.apply_series_generator() 852 853 # wrap results

~\anaconda3\envs\MovingPandas\lib\site-packages\pandas\core\apply.py in apply_series_generator(self) 865 for i, v in enumerate(series_gen): 866 # ignore SettingWithCopy here in case the user mutates --> 867 results[i] = self.f(v) 868 if isinstance(results[i], ABCSeries): 869 # If we have a view on v, we need to make a copy because

~\anaconda3\envs\MovingPandas\lib\site-packages\movingpandas\trajectory_generalizer.py in (rec) 253 254 dists = df.apply( --> 255 lambda rec: self._dist_from_calced( 256 rec, df.index.min().to_pydatetime(), df.geometry.iloc[0], de, dx, dy 257 ),

~\anaconda3\envs\MovingPandas\lib\site-packages\movingpandas\trajectory_generalizer.py in _dist_from_calced(self, rec, start_t, start_geom, de, dx, dy) 276 di = (rec.name - start_t).total_seconds() 277 calced = Point(start_geom.x + dx * di / de, start_geom.y + dy * di / de) --> 278 return rec.geometry.distance(calced)

~\anaconda3\envs\MovingPandas\lib\site-packages\pandas\core\generic.py in getattr(self, name) 5581 ): 5582 return self[name] -> 5583 return object.getattribute(self, name) 5584 5585 def setattr(self, name: str, value) -> None:

AttributeError: 'Series' object has no attribute 'geometry'`

Hi, first of all thanks a lot for this awesome project ! I've recently encountered a problem that I wanted to share.

Problem description

When working across the antimeridian, the interpolate_position_at function returns a unexpected value. This is due to the fact that the shapely.geometry Linestring object is considered as going across Greenwich meridian, despite the extremities of the line being very close.

One possible fix would be, if two positions are close and separated by the antimeridian, to add 360° to the negative longitude and subtract back the 360° to the interpolated longitude if it is above 180°.

Following on from https://github.com/anitagraser/movingpandas/pull/260 we could also remove matploblib as a dependency. For example, it's not a dependency in geopandas https://github.com/geopandas/geopandas/blob/main/pyproject.toml#L43

Seeing some doc failures with https://github.com/anitagraser/movingpandas/blob/main/docs/environment.yml e.g. https://readthedocs.org/projects/movingpandas/builds/18774152/

The pinning of versions in the doc env may be able to be removed.

Since geodesic distance measurements have replaced spherical ones, the parameter name now is misleading:

https://github.com/anitagraser/movingpandas/blob/81ebf4f749376f3d43ff330d7b70a25a6db7913a/movingpandas/geometry_utils.py#L77-L85