ctapipe Bi-Weekly Digest — Activity since 2026-01-17

» cta-observatory/ctapipe | cta-observatory/ctapipe

• cta-observatory/ctapipe repo issues [back to top]

  • Provenance crashes because it cannot serialise Astropy columns #2918 [Voutsi] bug

    Describe the bug Hello, while running the telescope cross calibration, I encountered the following error: Traceback...

    • kosack: I think this was missed in a previous PR review: astropy columns cannot be configuratin...

    • kosack: @Voutsi can you explain a bit better what the use case of changing the schema is? Is what I...

    • Voutsi: @kosack use case of changing the schema is I simply need a way to filter the dl2 data using...

    • kosack: To that end, I think the solution you are proposing in #2919 will work. Great, I think using...

  • shower dist histogram storage could lead to rounding errors #2924 [kosack] bug

    Describe the bug I want to soon implement support arbitrary shaped shower distributions, and noticed this when...

  • Allow specifying column metadata for features in FeatureGenerator #2922 [kosack] enhancement

    Please describe the use case that requires this feature. FeatureGenerator transforms an input astropy Table to an...

    • maxnoe: I would prefer the more expressive dict-base syntax, it's also easier to extend later with...

    • maxnoe: Edit: since inputs to the expressions have units, the output also will have one. So unit should...

    • kosack: Edit: since inputs to the expressions have units, the output also will have one. So unit should...

  • Numba caching race condition might lead to segfaults when importing ctapipe #2837 [maxnoe] bug

    Describe the bug A clear and concise description of what the bug is. @jsitarek and @Elisa-Visentin report segfaults...

    • STSpencer: I've ran an experiment on the DESY cluster with a clean pip install of ctapipe 0.29. I can...

  • Store true muon impact point on mirror for muon simulations #2929 [maxnoe] enhancement

    Please describe the use case that requires this feature. Checking the peformance of the muon fitting needs access to...

  • `get_border_pixel_mask` flags all pixels for staggered layouts #2931 [shahjahanIqbal] bug

    Describe the bug ctapipe.instrument.CameraGeometry.get_border_pixel_mask fails to identify border pixels for a...

    • maxnoe: @shahjahanIqbal This is only a symptom, the real issue is the automatic detection of the...

    • maxnoe: This is essentially a hex grid, but with square pixels. It should be relatively easy to...

    • maxnoe: You can also see this from this...

    • maxnoe: @shahjahanIqbal I opened a draft pull request here: #2932 that fixes this immediate issue. It...

  • Sonarqube shows missing coverage for definitely covered lines #2938 [maxnoe]

    https://sonar-ctao.zeuthen.desy.de/component_measures?id=cta-observatory_ctapipe_6122e87b-83f3-4db1-8287-457e752adf01...

    • maxnoe: Fixed by merging #2939

• cta-observatory/ctapipe repo pull_requests [back to top]

  • Prepare v0.28.1 #2904 [maxnoe] no-changelog-needed

    • maxnoe: Superseded by 0.29

  • First step of implementing CEP-002: Change container structure #2204 [maxnoe]

    • Tobychev

      • Tobychev: With the tel removed this container now only contains one thing, so its not clear what the extra...

        • maxnoe: There might be more per-subarray information in future than just the air shower...

  • Fix sphinx warnings #2923 [kosack]

    The documentation build is broken on all PRs right now, since sphinx 9.x was released and it causes some...

    • kosack: Docs building seems to be broken on sphinx=9 due to these linking issues, but works fine on...

    • maxnoe: Let's pin to sphinx 8 here and update in a decitated PR.

    • kosack: Pinning to 8.2 here doesn't seem to have worked. I'm not sure why. On my laptop, 8.2 worked...

    • maxnoe

  • Rename konrad weights to aspect-weighted-intensity weights #2880 [LukasBeiske]

    As discussed in the past, this renames the "konrad" weights for averaging image-wise ml predictions. Should we also...

    • kosack

    • maxnoe

  • Add a component to reduce waveform readout window #2780 [maxnoe]

    The question is now where to best apply this. I see a couple of options: Add directly to ctapipe-process, this...

    • mdpunch: Still seems fine! (Next step, to add a "GRB HaST" component in another pull request?)

  • Fix PendingDeprecationWarning for docutils arg #2860 [maxnoe] no-changelog-needed

    Fixes: src/ctapipe/core/tests/test_component.py::test_component_html_repr ...

    • kosack

    • mexanick

  • Merge telescope-wise data from the same OB #2916 [TjarkMiener]

    This PR adds a new merge strategy for combining telescope-wise data from the same OB. It also added a static method...

    • TjarkMiener: @Voutsi the default merge strategy will check if the subarrays match between different...

      • maxnoe: I think we before we continue here with implementation we need a better formulation of the...

    • maxnoe: @maxnoe , I would like to add here the use case of telescope cross calibration. There we want to...

    • TjarkMiener: @maxnoe Sorry for the long silence, I misinterpreted your comment about the multiple files...

    • maxnoe: Now, I think you mean that the cross calibration should just be able to digest per obs_id stereo...

    • maxnoe: DL1b per tel_id will be merged in each OB via ctapipe-merge with...

    • TjarkMiener: I have implemented the suggestions and discussion points above. This PR would be now ready for...

    • maxnoe: The docs build is fixed in main, please update

    • TjarkMiener: The docs build is fixed in main, please update Thanks @maxnoe it is passing now. Regarding the...

    • mexanick

    • maxnoe: A second motivation comes from CTLearn, where some stereo deep learning models operate on...

    • TjarkMiener: where we might create broken subarray descriptions and non-matching trigger and dl2 tables I...

  • Add mock for decorators of optional dependencies #2925 [maxnoe] no-changelog-needed

    As discussed in #2920

  • Refactor DL2EventLoader to use FeatureGenerator #2919 [kosack]

    This fixes #2918 , and was triggered by it. Configuration variables must be traitlets, not complex classes without...

    • kosack: Looking deeper at this code, it's somewhat fragile, and confusing. I think it may need some...

    • kosack: I'm also thinking a bit forward to a missing interface: event-wise algorithm selection, so I...

    • maxnoe: Yes, this should never have been a traitlet of type List(Column), at least not without defining...

    • kosack: Actually I have an even better API than my suggestion above: what the DL2EventPreprocessor does ...

    • LukasBeiske: Actually I have an even better API than my suggestion above: what the DL2EventPreprocessor does...

    • kosack: I've currently replaced all the genearted columns in the old implementation with...

    • maxnoe: By allowing passing in a callable and restricting to UseEnum with a default here, we end up in...

      • kosack: Good point. I could just restrict this to pre-defined spectra and not allow an arbitrary one to...

      • maxnoe: Instead of the enum, we could make traitlets wrappers for the spectra themselves. spectrum_cls:...

      • maxnoe: to still allow setting pre-configured ones, you could add a spectrum_name with default None that...

      • kosack: That would be a nice separate PR I think. I'll open an issue. For now I'll just restrict to the...

    • maxnoe: We need a fov_offset_min here, e.g. to compute sensitivity in fov offset bands.

      • kosack: That was what is in the RadialEventWeighter implementation already,but I suppose this Simple...

    • maxnoe: - [true_fov_offset, "angular_separation(reco_fov_lon, reco_fov_lat, 0*u.deg, 0*u.deg)" -...

    • maxnoe: Is spectially actually a commonly used word for this?

      • kosack: No, fixed. Not sure why the spell checker in pre-commit didn't catch that though!

    • kosack: In refactoring this, I think we no longer really need DL2EventLoader at all. The...

    • kosack: Since this refactoring is getting large, and I don't want to break all the working code, I will...

  • Make HTML reprs for Components recursive #2926 [kosack]

    Minor cosmetic change that I found useful when debugging complex Components that contain other components: recurse...

    • maxnoe: Suggested change the Jupyter...

    • maxnoe: This doesn't actually print the attributes name, so all I know is that there is e.g. a...

      • kosack: done.

    • maxnoe

  • Fix FeatureGenerator gives wrong units #2921 [kosack]

    FeatureGeneartors currently support both Table and QTable as inputs, and both are tested. However, unit propagation...

    • kosack: Do we rely anywhere on using FeatureGenerator on Tables with unit columns? If not, I can modify...

    • maxnoe: We should just add QTable(table, copy=False) to the call function I think. That's a very cheap operation.

    • maxnoe: We should just change the method above: ...

      • kosack: I think it's clearer to do it explicitly here - that function may be useful elsewhere, and...

      • maxnoe: It think you misunderstood. The code as it is here now First makes a deepcopy of the input...

      • kosack: I missed that there was a deepcopy done in the conversion. I assumed QTable(table) would return...

    • kosack: Once #2923 is merged, the docs here should pass.

    • maxnoe: Docs build is fixed in main, please rebase

    • maxnoe: Suggested change def...

    • maxnoe

    • mexanick

  • Minimal initial commit of FreePACT code #2920 [ParsonsRD]

    This is a commit of a simple implementation of the FreePACT code. Updates to the ImPACT code are included from #2705

    • maxnoe: I'd add a new extra here called impact or freepact that depnds on tensorflow and iminuit (you...

    • maxnoe: It seems current tensorflow does not yet support python 3.14. You can remove the impact...

    • maxnoe: I.e. after adding the impact extra, you add it in the list here in the last line to include it...

    • ParsonsRD: @maxnoe I'm a bit lost what to do here, I can add the tensorflow dependencies for the older...

    • maxnoe: Please do not change the python requirement here.

      • ParsonsRD: Yep this was not me, I must have messed up the merge

    • maxnoe: This must be some broken merge conflict or rebase resolution, you are reverting some recent...

      • ParsonsRD: This must be a broken merge. Not sure why that happened, I'll revert it.

    • maxnoe: @ParsonsRD you need to make the import of tensorflow actually optional in the code. I.e. you...

    • ParsonsRD: Ugh sorry yes that should have been obvious. I blame on being very tired today :-).

    • maxnoe: Please leave iminuit in all, impact is not the only thing that needs it.

    • maxnoe: The network interpolator still throws on import: File...

    • ParsonsRD: OK I can merge it, yeah I think I'm missing something here. The run of ctapipe-info in the CI...

    • maxnoe: I guess with what you suggested it would not throw this error unless I actually declare an...

    • maxnoe: No need to override if you just call super().__init__()

    • maxnoe: Leave one empty line between first line of docstring and the rest. No need to repeat the...

    • maxnoe: iminuit should be here as well, so that people who do pip install ctapipe[impact] get it. It...

    • maxnoe: Here is a working complete example also covering possibly passing arguments to the tf.function...

    • maxnoe: I opened a separate PR here: #2925

    • ParsonsRD: tf and keras does not seem to play nice with the latest version of numpy. But I can't get my...

    • maxnoe: tf and keras does not seem to play nice with the latest version of numpy. But I can't get my...

    • maxnoe: For the warnings that come from importing protobuf via tensorflow, I have these in another...

    • maxnoe: you need one more colon here. It is <action>:<regex for...

      • ParsonsRD: Thanks, I couldn't find an explanation of this anywhere and couldn't find a way of testing it...

      • maxnoe: These warnings are triggered in the oldest-deps build, where we test against the oldest still...

      • maxnoe: In this case, it's matplotlib=3.8 raising these warnings.

      • maxnoe: If you don't mind, I'll push a commit to update the pyproject

    • maxnoe: We use the bibtex extension for sphinx, please add the reference to the docs/references.bib file...

    • maxnoe: Please remove the impact extra from the CI job for minimal-dependencies, the task of this job is...

    • maxnoe: Please remove the impact extra from the CI job for minimal-dependencies, the task of this job is...

    • maxnoe: The tests are passing now. You do not need to worry about the pending check, that comes from...

    • ParsonsRD: OK great, thanks for the help in working through the CI issues. I will work through the...

  • Add full cut optimization as introduced in pyirf 0.13 #2789 [LukasBeiske]

    This changes the PointSourceSensitivityOptimizer to use the full cut optimization introduced in pyirf 0.13. The...

  • Fits monitoring source #2933 [mexanick] new functionality module:IO

  • set is_valid to true in VarianceExtractor #2930 [TjarkMiener] bug

    only set false if all values are NaNs currently all is_valid flags are set to the default container field which is false.

    • maxnoe: Suggested change ...

    • maxnoe: needs two backticks in rst: Suggested change ...

    • kosack

    • maxnoe

  • Feature: EventWeighter #2927 [kosack]

    This is part of the DL2 code refactoring (see #2919). It introduces a new Component hierachy...

    • maxnoe: What is the use case for this? I am opposed to adding classes that throw not implemented error.

      • kosack: For the future when we make IRFs with offset and phi dependence. I can remove it for now, but...

      • maxnoe: Why would you need different weights per phi bin for that?

      • kosack: Actually, it was pretty simple to just add an implementation, so it's there now.

      • maxnoe: Please explain why we would actually need it.

      • kosack: For simulations, it's not particularly useful since they are always radially symmetic in true...

      • kosack: Anyhow, I agree I shouldn't add things I don't need yet, so removed.

    • maxnoe: Let's use...

      • maxnoe: and we should also store the "valid" column.

    • maxnoe: Wouldn't this example be better placed in the docstring of the event weighter instead of the changelog?

      • kosack: Yes, though it's ok to keep it here as It hink the changelog is nice for advertising new...

  • Feature: EventPreprocessor #2928 [kosack]

    Introduces an EventPreprocessor class that can generically transform an event table by applying the following...

    • maxnoe: It does not convert to NominalFrame. It converts to GADF FoV coordinates, since you added a...

      • kosack: good point - or maybe it's better to just remove the minus sign, and apply that correction it in...

      • kosack: I just removed the minus, as I think it's not needed. It's preferred to apply it only at the end...

    • maxnoe: test is not updated.

    • maxnoe: comment is now obsolete

    • maxnoe: Why is the conversion needed? Shouldn't UseEnum take care of that? ...

      • kosack: done. leftover from an older imlpementation before I switched to UseEnum

    • maxnoe: Please still pass parent, otherwise the logging and configuration export won't be setup properly.

    • maxnoe: Please still pass parent, otherwise the logging and configuration export won't be setup properly.

  • Provide better error in case of missing changelog snippet #2934 [maxnoe] no-changelog-needed

    • maxnoe: The error is now:...

    • mexanick

  • Add a StereoDispCombiner #2731 [Hckjs] enhancement algorithm module:reco

    This is basically the StereoCombiner already implemented here in magic-cta-pipe. It combines the mono DISP...

    • maxnoe: Does it make sense to make this configurable? Shouldn't it just be GEOMETRY?

      • maxnoe: I'm not sure I understand the point. This stereo combiner can only ever reconstruct geometry. So...

      • Hckjs: ctapipe/src/ctapipe/reco/sklearn.py Lines 601 to 605 in b2dc9c2 ...

    • maxnoe: Instead of using this sign score limit, I think EventDisplay requires a minimum angle between...

      • Hckjs: Just to be clear: Do they check the angle just for events with multiplicity of two? Or are...

      • Hckjs: https://github.com/Eventdisplay/Eventdisplay/blob/56dfa2f3c0b1b3fe359ac4b1d5edfe7045b25765/src/VD...

    • maxnoe: "konrad" was renamed recently, see #2880

    • maxnoe: You assume you get a ImageParametersContainer here, right? Should check for this explicitly

  • Not store tel-wise fixed pointing tables if telescope_events is set to false #2936 [TjarkMiener] bug

    • maxnoe

  • Retry vizier queries #2935 [maxnoe] no-changelog-needed

    We observed since some time that CI pipelines fail from time to time due to vizier being unavailable. I'm not sure...

    • maxnoe: The issue is not vizier being unavailable, but actually a rate limit. So retrying would be even worse.

    • maxnoe: https://docs.visier.com/developer/apis/rate-limit.htm

  • Minimize VIZIER queries in CI by only running catalog tests in coverage job #2937 [maxnoe] no-changelog-needed

  • Fix sonar #2939 [maxnoe] no-changelog-needed

  • Implement square pixels on a hexgrid #2932 [maxnoe]

    Store new field in geometry files tests import numpy as np import matplotlib.pyplot as plt from ctapipe.instrument...

    • AtreyeeS: Thanks a lot for the quick fix @maxnoe ! This addresses our requirement for now

created at 2026-01-31 06:48 by dinghy 1.4.1.