As an experiment, we are trying to set up a rotating duty to take a first look at all contributions (PRs and issues) from the community, label them and ping relevant reviewers.
Scripts for generating the reports are available in this repository.
mar/2020
apr/2020
may/2020
jun/2020
jul/2020
ago/2020
sep/2020
oct/2020
nov/2020
dec/2020
The script to generate the report below is on https://gist.github.com/emmanuelle/6937f6ad426754d40f5e9bcfdebd013d For now it's only a gist, you need to put your own Github token in the script. There is room for improvement in the script, in particular I use aggressive thresholds to cap the number of queried pull requests and issues (because it takes a lot of time) but we might miss some updates about very old PRs and issues.
Some other comments
out vs inplace Many thanks to all reviewers!!
data module don't work [Windows]data module don't work [Windows]doc/examples/registration are not displayed in the gallerymultichannel= in favour of channel_axis=multichannel parameter handlingHi all!
Thanks to @emmanuelle this issue is going to get really long. =P See the output of her script below. I didn't reach her very high standard of merging every PR, but I reckon that's because people are being nitpicky about my brilliant #4585. =P Thanks @emmanuelle in particular for getting our CI green again, that's just excellent.
Could @scikit-image/core comment on the behaviour change picked up by @lagru in: https://github.com/scikit-image/scikit-image/pull/4585/files#r411019161
Thanks!
separate_stains docstringrescale_intensityA good activity this week. An issue #4603 raised again the question how to decide the min version of requirements. @soupault opened a great issue (https://github.com/scikit-image/scikit-image/issues/4608) summarizing our current decisions and proposing policies for the different cases. This PR requires attention of @scikit-image/core in particular.
Kudos again to @soupault who fixed a ton of broken links in our code: https://github.com/scikit-image/scikit-image/pull/4600
@emmanuelle put the script to generate this report in a repository, so we can easily submit PRs.
I tried to fix an issue in skeletonize3D based on a patch that have been submitted to the matlab implementation of this algorithm. I might not have fully understood the matlab code, but the modifications seem to be too drastic and lead to another incorrect behavior. Fiji still have the bug. I have not done the work to read carefully the paper to check the current implementation regarding the original algorithm.
I also investigated the loading page performances of our gallery. It turns out that the page is slow due to thumbnails. We merged a PR to reduce the resolution of thumbnails, which decreases by a factor 2 the image sizes, while keeping a correct rendering. I opened an issue on the sphinx-gallery tracker to raise this problem. Matplotlib and scikit-learn are also interested in better performances. One sphinx-gallery dev proposed a PR to use optipng (after @GaelVaroquaux's recommendations). This has to be tested.
I hope I did not missed any other important point.
exposure.rescale_intensity for constant input.equalize_adapthist implementation (CLAHE) to N dimensionsexposure.rescale_intensity for constant input.equalize_adapthist border artifacts still presentexposure.equalize_adapthist can fail for constant inputs due to incorrect behavior in rescale_intensityseparate_stains docstringexposure.equalize_adapthist can fail for constant inputs due to incorrect behavior in rescale_intensityrescale_intensityThank you @alexdesiqueira who did most of the work. Our CI is still failing because of Cython pre-release, I hope that they have a new pre-release coming soon because it's not nice to have a red CI. I hope that we can release 0.17 this week, I will take care of it with the help of @jni.
equalize_adapthist implementation (CLAHE) to N dimensionsequalize_adapthist border artifacts still presentAs you know last week was a release week and we're still in release-land with 0.17.2 coming up this week (meaning some backports of bug fixes for bugs introduced in 0.17). Apart from release stuff it'd be great to take a look at #3226 (pythran demo)
equalize_adapthist: More efficient processing scheme with symmetrical paddingdtype.convertplane, row, col notation in all docstringsplane, row, col dimensions ordering as specified in the skimage conventions pageseparate_stains docstringLots of new documentation and small fixes proposed, some backports and a discussion on our install docs — it would be interesting to have some thoughts from all @scikit-image/core to find a consensus and work on that quickly.
dtype.convertdtype.convertThis seemed like a productive week to me. Some highlights:
The gallery examples mentioned above and #4717 as well as #4734 might be good places to review if you find the time. And we still need someone to take over the community management this week (2020/05/25 - 2020/05/31)!
none
..versionchanged:: tags to all API changes from 0.17Last week looks like a quite week compared to the previous weeks, but as the following highlights shows, things are going very well:
trainable_segmentation method which is a good candidate for integration in skimage.segmentation,peak_local_max but waits for some review,There is no one in charge for the community management this week, we are still looking for a last minute volunteer ;)
peak_local_max takes a very long time to completeNot too much activity this week, but several issues related to Mac/Win were described.
none
none
This week we merged an awesome biology example and fixed random_walker.
none
Last week highlights:
none
An average level of activity last week I'd say. The CI is currently broken because of matplotlib pre-release which is annoying, I opened #4805 to discuss our general CI process related to pre-releases. Several open PRs have a good momentum but need some input, for example #4767, #4794, #4739, #4444.
Hello all! And hello @alexdesiqueira, thanks for picking SciPy week to be on call, a brave move! :sweat_smile:
Here's the summary for last week from @emmanuelle's script. Note that I've modified it to do Monday to Monday at midnight UTC, rather than at whatever time the current maintainer decides to run it. (I'll submit a PR for this tomorrow.) This will make sure that issues etc are always caught by a run of the script.
I'm also going to curate the output to display PRs and issues opened and closed in the same week as crossed out, which should give a better sense of how we are doing on response time. And I'll comment on PRs/issues inline below.
Long term, I want to modify the script to produce some data analysis about how we are doing. Response time, sentiment, etc. Ambitious perhaps but I think it would be both motivating and informative!
Ok enough planning:
none
equivalent_diameter -> equivalent_diameter_area (:point_left: needs discussion)feret_diameter -> max_feret_diameter ? (:point_left: needs discussion)Going through these notes manually, rather than just posting the output, led me to merge two PRs that might have just sat there for months. So it's a really good exercise to go through these slowly at the end of a shift!
Also, thanks to those two merges, we merged 8 PRs this week but 7 were opened, so we are net negative for PRs! :tada: For issues we are +5 so there is some work to be done! :grimacing:
Again, good luck @alexdesiqueira this week! :blush:
Lots of work this week due to our SciPy 2020 sprint. Again, we'd like to thank everyone for your time, discussions and contributions!
equalize_adapthist: More efficient processing scheme with symmetrical paddingskimage.measure.find_contoursequalize_adapthist returns input image for clip value of 1.0multichannel= in favour of channel_axis=multichannel parameter handlingferet_diameter -> max_feret_diameter ?@jni:
Going through these notes manually, rather than just posting the output, led me to merge two PRs that might have just sat there for months. So it's a really good exercise to go through these slowly at the end of a shift!
This is a great idea! I tried that in the past, but I couldn't do it effectively yet. I'll try to add it to my process as well.
It was the week just after the scipy sprint, so quite a lot of activity. #4844, #4847, #4850, #4852 and #4853 are open PRs resulting from the sprint, so let's keep an eye on them (I think #4844 and #4852 can be merged quite soon). #4850 needs a review (3D tutorial).
Our CI is currently broken due to a matplotlib bug with PYTHONOPTIMIZE=2 (see #4854).
There have been interesting discussions in #4794 about type annotations, I'm thinking of organizing a dev meeting which would discuss specifically this topic.
Please also take a look at the list of "Older pull requests with new comments or commits", some of them would need a round of review.
feret_diameter -> max_feret_diameter ?This week we merged the great 3D AHE example and some documentation on how to contribute with a gallery example. Thanks @m-albert @mkcor!
none
equalize_adapthist: More efficient processing scheme with symmetrical paddingnone
Nice features of the week. We now have doc builds for PRs by @emmanuelle . A 3D dataset has been contributed this week by @GenevieveBuckley which is extremely valuable. I noted also the nice ongoing contribution #4851 by @FirefoxMetzger on background substraction, which received a lot of comments already and for which we must converge anytime soon.
We also have several PRs/issues for documentation and performance issues.
draw.circleQuite a busy week for mid-August standard. Alex is introducing a new eagle image and replacing the camera man image by a new one due to license issues. This change might cause some tests to fail, so please pay attention to #4913.
For the coming week we don't have anybody yet registered to be in charge, and same for the following weeks. For this week I can follow up but only until Thursday and then I'm on vacation for 10 days, so we'd need someone else to take over starting from Friday (or for the whole week depending on their availability).
hough_line_peaks does not respect min_distance argumentdraw.circleLots of infrastructure work this week. Also, data/eagle.png is now available :tada:
none
width and height swapped in morphology.rectangleimage description in docstring: towards uniformity?viewer and skiviwidth and height swapped in morphology.rectangleLast week was quite productive with two bug fixes merged! Thanks to @sciunto we also now deprecate skivi :tada: I would be very happy if we keep moving on this way and soon deprecate skimage.io and skimage.viewer :wink:.
I just finally want to bring your attention to a decision to be taken concerning edge cases of peak_local_max
filled_area in measure.regionprop wrong? -> regionprop fails to fill a shapehough_line_peaks does not respect min_distance argumentskimage.measure.find_contoursviewer and skiviDuring this first week of september, the activity was lower than usual. We had a broken CI for two reasons. With @hmaarrfk we worked on fixing it. There is still issues with OSX jobs on travis caused by a timeout. We added a verbose flag to the command, which helps from time to time, but is not always sufficient. We were unsuccessful applying correctly travis_wait, we must try again. See #4955
I noted activities around PRs on 3D data sets by @mkcor and @hmaarrfk
I also made triage on old issues.
filled_area is wronghough_line_peaks does not respect min_distance argumentviewer and skiviHi @alexdesiqueira! Here's the report for this week. =) We're -2 on PRs (:tada:) but +1 on issues (:disappointed:). Pretty good week though. :blush:
Note: I manually edited the report below to mark the PRs that were opened and closed this week.
none
I'd consider not too much movement this week (although I was taking care of #4913, so didn't see too much outside that :slightly_smiling_face: ). We had our Atlantic meeting today — with @emmanuelle @jni @grlee77 @lagru @sciunto and myself — preparing for 0.18. Great to see you all!
none
Not a lot of new PRs this week, but I think this is because the core team is busy trying to merge existing ones (the "Older pull requests with new comments or commits") for the 0.18 milestone https://github.com/scikit-image/scikit-image/pulls?q=is%3Aopen+is%3Apr+milestone%3A0.18
We need someone for next week!
This has been my first week as the person in charge, yay! I'm not sure I've followed everything as closely as I should have; the beginning of the week was pretty quiet.
PS: Just so you know, the utility script for this weekly update spit out "2020/9/28 - 2020/10/5" as the title for this post, and I shamelessly edited to "2020/9/28 - 2020/10/4" (meant to fix the inconsistency, but then I ran out of GitHub API pokes, even though I generated a token, but didn't export it, and anyway... next time). :wink:
none
Some movement this week. I'll point to #4898, which will help a lot when we release 0.18. #4739 needs some reviews; please check it out if you can.
Last week was in positive balance with 6 new PR VS 6 merged PR and 4 closed issues VS 3 new ones :tada:. I just want to draw your attention to #5020, #5006, #4760 and #4739 for some review, thank you all :wink:.
Thank you @rfezzani! Gotta love paying down that debt! :tada:
I notice another gap in the calendar. I'll take this week. Been wanting to make another push on 0.18! :crossed_fingers:
Hi everyone,
Last week was pretty quiet but we still managed to accrue a negative balance. :disappointed: I made a push on #4444 (thank you @emmanuelle for the help there) as well as some issues related to 0.18.
As always, I'd like to point everyone to the list of PRs with one approved review, to see if we can merge some of these soon!
none
Please keep an eye on #5053 y'all; we're dropping support to py3.6 and improving machinery for py3.8.
shape=None and shape=actual_shapeHello! Since they don't appear below, maybe I should draw our attention to PRs #2815 and #3070 ? Both are in for 0.18 and have one approval at the moment.
none
Oups! I am so sorry, I just realized that it was my turn last week :disappointed:
Oups! I am so sorry, I just realized that it was my turn last week 😞
you will change that emoji to something glad if you realize that your effort will happy a kid in paint by python app and or continue to rescue a father from cancer die by scikit scientific image recognition :)
The last week has mostly been work related to preparing the 0.18 release. In the last few weeks, CI and wheel building jobs were moved off of Travis-CI now that free credits are much more limited. We may keep ARM wheel builds on Travis, but all x86 jobs have been moved elsewhere.
A new proposed feature is installing scikit-image[data] instead of just scikit-image to get an install with the optional pooch dependency included. The lazy importing proposed in #5101 looks like a nice convenience.
Also, there is a request for comments on imagio API imageio/imageio#569 (related to IO offloading issue #5036)
none
Lots of movement this week as we (as in @jni :slightly_smiling_face:) prepare our 0.18 release! Thank you all for your hard work!
Most helpful comment
2020/4/20 - 2020/4/27
A good activity this week. An issue #4603 raised again the question how to decide the min version of requirements. @soupault opened a great issue (https://github.com/scikit-image/scikit-image/issues/4608) summarizing our current decisions and proposing policies for the different cases. This PR requires attention of @scikit-image/core in particular.
Kudos again to @soupault who fixed a ton of broken links in our code: https://github.com/scikit-image/scikit-image/pull/4600
@emmanuelle put the script to generate this report in a repository, so we can easily submit PRs.
I tried to fix an issue in skeletonize3D based on a patch that have been submitted to the matlab implementation of this algorithm. I might not have fully understood the matlab code, but the modifications seem to be too drastic and lead to another incorrect behavior. Fiji still have the bug. I have not done the work to read carefully the paper to check the current implementation regarding the original algorithm.
I also investigated the loading page performances of our gallery. It turns out that the page is slow due to thumbnails. We merged a PR to reduce the resolution of thumbnails, which decreases by a factor 2 the image sizes, while keeping a correct rendering. I opened an issue on the sphinx-gallery tracker to raise this problem. Matplotlib and scikit-learn are also interested in better performances. One sphinx-gallery dev proposed a PR to use optipng (after @GaelVaroquaux's recommendations). This has to be tested.
I hope I did not missed any other important point.
2020/4/20 - 2020/4/27
New pull requests open last week
4620 Add .mypy_cache to the gitignore
4619 added one example to flood fill tutorial
4617 Change math formatting in the docstrings
4615 Fixed wrong behaviour of
exposure.rescale_intensityfor constant input.4614 Change label2rgb default background value from -1 to 0
4613 Improve loading speed of our gallery by reducing the thumbnail size
4612 Allow very large values in relabel_sequential
4610 Add notes on deprecation warnings in marching_cube_* and gray2rgb
4609 Fix register_translation warning message
4607 Clarify error message when montaging multichannel nD images and multichannel=False
4604 Update min required sphinx version for sphinx-copybutton
4602 Fix typo in cython import_array
4600 Fix broken and permanently moved links
4599 Add example of natsort usage in documentation
4598 Extended existing
equalize_adapthistimplementation (CLAHE) to N dimensionsClosed pull requests (not merged)
4417 Fix rgb2gray docstring to show that 2-D arrays are allowed as inputs
Merged pull requests
4620 Add .mypy_cache to the gitignore
4617 Change math formatting in the docstrings
4615 Fixed wrong behaviour of
exposure.rescale_intensityfor constant input.4613 Improve loading speed of our gallery by reducing the thumbnail size
4610 Add notes on deprecation warnings in marching_cube_* and gray2rgb
4609 Fix register_translation warning message
4607 Clarify error message when montaging multichannel nD images and multichannel=False
4604 Update min required sphinx version for sphinx-copybutton
4602 Fix typo in cython import_array
4600 Fix broken and permanently moved links
4599 Add example of natsort usage in documentation
4590 Rename convert to _convert, as it is a private function
4585 Update rescale_intensity to prevent under/overflow and produce proper output dtype
4528 Add example code for generating structuring elements.
4502 Move register_translation from skimage.feature to skimage.registration
4496 Fix h_maxima/minima strange behaviors on floating point image input
4467 Complete deprecation of circle in morphsnakes
Older pull requests with new comments or commits
4531 Fixed automatic color generation in label2rgb function
4517 WIP: stub sphinx example to illustrate data type relevance
4515 example: register rotation and scaling with no shared center
4503 Move masked_register_translation from feature to registration
4435 3D Rank Filters WIP
new issues
4621 Does Difference of Gaussian need normalization?
4618 flood_fill and flood are both in segmentation and morphology
4616 Search links broken in dev documentation
4611 Can't load most images from data package
4608 RFC: Support cycle for the upstream and the downstream dependencies
4606 Misleading "ValueError: Input array has to be either 3- or 4-dimensional" in montage
4605 Check in CI whether we can bump the versions of the requirements
4603 Documentation build failure with sphinx 2.4
4601 Discuss gallery structure to bring out bioimaging examples.
4597
equalize_adapthistborder artifacts still present4596
exposure.equalize_adapthistcan fail for constant inputs due to incorrect behavior inrescale_intensityOlder issues updated last week
4594 Broken link in
separate_stainsdocstring4544 Simplify the release process
4507 Automatic color generation in the label2rgb function
4486 Calendar of community management
3015 WIP: Turn off relabelling in label2rgb
2761 Generalised existing CLAHE implementation to support n dimensions
Closed issues
4611 Can't load most images from data package
4603 Documentation build failure with sphinx 2.4
4596
exposure.equalize_adapthistcan fail for constant inputs due to incorrect behavior inrescale_intensity4595 Broken links in docs
4583 funky behavior of
rescale_intensity4513 Add example of natsort usage in documentation
4431 Add examples to docstring for structuring elements to visualize the shape
4194 h_maxima/minima strange behaviors on floating point image input
3014 Magic dimensionality guessing in gray2rgb