satellogic / open-impact Goto Github PK

The Stanford hyperspectral notebook cannot be run because the user lacks the necessary permissions to download the image.

I had to make these changes to environment.yml to make it work on Linux:

diff --git a/environment.yml b/environment.yml
index 454b974..dd853dd 100644
--- a/environment.yml
+++ b/environment.yml
@@ -4,7 +4,6 @@ channels:
 - defaults
 dependencies:
 - affine=2.2.0=py_0
-- appnope=0.1.0=py36_0
 - asn1crypto=0.24.0=py36_0
 - attrs=17.4.0=py_0
 - backports=1.0=py36_1
@@ -120,7 +119,7 @@ dependencies:
 - pytz=2018.3=py_0
 - pywavelets=0.5.2=py36_1
 - pyzmq=17.0.0=py36_3
-- qt=5.6.2=h9e3eb04_4
+- qt=5.6.2
 - qtconsole=4.3.1=py36_0
 - rasterio=0.36.0=py36_3
 - readline=7.0=0
@@ -151,12 +150,9 @@ dependencies:
 - xz=5.2.3=0
 - zeromq=4.2.3=2
 - zlib=1.2.11=0
-- anaconda=custom=py36ha4fed55_0
 - intel-openmp=2018.0.0=8
-- libgfortran=3.0.1=h93005f0_2
-- mkl=2018.0.2=1
-- numpy=1.14.2=py36ha9ae307_0
-- scipy=1.0.0=py36h1de22e9_0
+- numpy=1.14.2
+- scipy=1.0.0
 - pip:
   - backports.functools-lru-cache==1.5
   - dask==0.17.2

The reason is that, since the introduction of newer conda versions, environment.yml files are no longer cross platform by default:

conda/conda#6220

(Other related issues: conda/conda#1033 ContinuumIO/anaconda-issues#546)

The solution is to export the environment with the --no-builds flag, which needs conda >= 4.3.31 to work. However, be careful with which conda version you use because now conda-forge is in an "everything is broken" state which hopefully will go away soon, but not before AnacondaCON conda-forge/conda-feedstock#43

So my advice is to use this in the meanwhile:

$ conda install "conda<4.4" -c conda-forge

As a last note, there's an open issue about pip packages in environment.yml files conda/conda#5901 that might end up in broken installations, so my advice is to remove them from the environment.yml file and tell the users to install them manually after creating the environment.

Nobody said it would be easy :)

MAke sure this is replicable, scripting the creation of the same conda environment, and documentation.

Given a hyperpixel, compare it with data from other instruments, like micro, or Landsat.

Switch over to download the data using Telluric, not the web explorer version.

Across notebooks, we use the same basic tools for visualization, download, exploring a hypercube...

It is probably a good DRY idea to collect all these into a single satellogic_utils.py file and then import satellogic_utils.

That probably also means a minimal documentation for the inputs of each function too.