-
Notifications
You must be signed in to change notification settings - Fork 10
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add tutorial notebook for TIKE #149
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #149 +/- ##
==========================================
- Coverage 93.92% 91.20% -2.73%
==========================================
Files 37 41 +4
Lines 1598 2149 +551
==========================================
+ Hits 1501 1960 +459
- Misses 97 189 +92 ☔ View full report in Codecov by Sentry. |
"source": [ | ||
"## Loading light curves into LCviz\n", | ||
"\n", | ||
"Because TIKE already runs in the cloud, it is fastest and easiest to load data that is also hosted in the cloud using the `s3fs` package, which allows you to access cloud data as if it were local. Here we will use [astroquery](https://astroquery.readthedocs.io/en/latest/) to get the URI for the file that we want to load. For more information about the code below, see [this TIKE webinar notebook](https://github.com/spacetelescope/tike_content/blob/main/content/notebooks/webinar-series/01-lightcurves/01-Lightcurves.ipynb).\n" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
since this mentions TIKE specifically, maybe we don't want to actually merge this PR here?
"light_curve = lightkurve.read(lc_uri)\n", | ||
"lcviz.load_data(light_curve)" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@bmorris3 - what would it take to be able to extend the upstream URI capability to be able to pass lc_uri
to load_data
directly?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It would require one more little block of logic in the jdaviz URI parser. That's work we need to do for RSP soon anyway. In their notebooks, they use this pattern on AWS:
import s3fs
import roman_datamodels as rdm
asdf_dir_uri = 's3://roman-sci-test-data-prod-summer-beta-test/'
fs = s3fs.S3FileSystem()
asdf_file_uri = asdf_dir_uri + 'ROMANISIM/DENSE_REGION/R0.5_DP0.5_PA0/r0000101001001001001_01101_0001_WFI16_cal.asdf'
with fs.open(asdf_file_uri, 'rb') as f:
file = rdm.open(f)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've tried running the notebook as-is on TIKE and it doesn't work yet, for reasons I don't understand. One challenge that we're likely to face is that lightkurve has tools for initializing a LightCurve
from a file path to a FITS file, but not from an open file stream to an HDU list, which is what fsspec
is likely to give us.
notebooks/tike_lcviz_tutorial.ipynb
Outdated
"# get the origin of the time axis in LCviz:\n", | ||
"time_coordinates = lcviz.app.data_collection[0].coords\n", | ||
"reference_time = time_coordinates.reference_time\n", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this isn't technically public API... can this be pulled out of lcviz.get_data(...).meta
instead?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Probably, I just pulled this from the LCvizExample notebook. Should probably change it there too if we don't want people using this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok, I'd vote to do that if possible to avoid publicizing any internal API if there are alternatives
notebooks/tike_lcviz_tutorial.ipynb
Outdated
"reference_time = time_coordinates.reference_time\n", | ||
"\n", | ||
"# literature ephemeris for hot Neptune planet HAT-P-11 b:\n", | ||
"morris2017_epoch = Time(2454605.89146, format='jd')\n", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
can we avoid using Time
entirely or is this needed because reference time is defined differently? Or is it good to show how to do this here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Again, I just pulled this from the example notebook, which I think @bmorris3 wrote. Probably a question for him.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why would you want to avoid using Time
? Time coordinates are dangerously ambiguous otherwise.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
because currently it needs to be passed to lcviz just as a float anyways, so creating a time object just to do to_value
seems to add complication to the notebook. Maybe we just link to docs on Time (whether or not we use it), since we don't want this to be a tutorial on that whole topic 🤷♂️
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm working through the notebook, and I think as long as the time system is written in the comment, I'm ok with avoiding astropy.time.Time
here. The most likely "gotcha" to worry about is that the units of the reference epoch, the period, and the time axis of the light curve must all be in the same units. If you had period in hours, this wouldn't work correctly. For this example, we're fine without specifying.
"\n", | ||
"- [LCviz documentation](https://lcviz.readthedocs.io/en/stable/index.html)\n", | ||
"- [LCviz Github](https://github.com/spacetelescope/lcviz)\n", | ||
"- [lightkurve documentation](https://lightkurve.github.io/lightkurve/)\n", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not sure if this link will break in the future when they fix docs.lightkurve.org
🤔
Co-authored-by: Kyle Conroy <[email protected]>
Co-authored-by: Kyle Conroy <[email protected]>
Co-authored-by: Kyle Conroy <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Pending support from Thomas to make the file loading work, this looks good!
notebooks/tike_lcviz_tutorial.ipynb
Outdated
"reference_time = time_coordinates.reference_time\n", | ||
"\n", | ||
"# literature ephemeris for hot Neptune planet HAT-P-11 b:\n", | ||
"morris2017_epoch = Time(2454605.89146, format='jd')\n", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm working through the notebook, and I think as long as the time system is written in the comment, I'm ok with avoiding astropy.time.Time
here. The most likely "gotcha" to worry about is that the units of the reference epoch, the period, and the time axis of the light curve must all be in the same units. If you had period in hours, this wouldn't work correctly. For this example, we're fine without specifying.
"* [Citing `astropy`](https://www.astropy.org/acknowledging.html)\n", | ||
"* [Citing `lightkurve`](http://docs.lightkurve.org/about/citing.html)\n", | ||
"\n", | ||
"And cite Jdaviz through its [Zenodo record](https://doi.org/10.5281/zenodo.5513927).\n" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@kecnry – should we make a zenodo record for lcviz? (I vote yes!)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, let's plan to do that with 1.0 release. I'll add it to the ticket.
} | ||
}, | ||
"source": [ | ||
"## Imports\n", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lcviz
isn't currently installed on the production version of TIKE – I'm assuming that will be done when this notebook is available?
Also, these imports only work if you're in the kernel named TESS Environment
. Should we mention that here?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Correct, they will install it once 1.0 is released AND they have this notebook. I think the general instructions handle selecting the kernel environment, but that's a good question for @ttdu
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lcviz will be installed in ~December, when we release a new environment for AAS.
we normally specify the default kernel in the notebook metadata, so that TESS Environment
is automatically selected. it never hurts to be verbose, but Kyle is correct that the general instructions cover this
"light_curve = lightkurve.read(lc_uri)\n", | ||
"lcviz.load_data(light_curve)" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've tried running the notebook as-is on TIKE and it doesn't work yet, for reasons I don't understand. One challenge that we're likely to face is that lightkurve has tools for initializing a LightCurve
from a file path to a FITS file, but not from an open file stream to an HDU list, which is what fsspec
is likely to give us.
Co-authored-by: Brett M. Morris <[email protected]>
moved to spacetelescope/tike_content#46 |
I figured I would open a PR here first to get comments, and then we can move/copy the notebook to wherever TIKE needs it for review.