how I set up an analysis repository building on a dataset (IN PROGRESS) #251
Replies: 3 comments 5 replies
-
instructions followed & one typo fixed. |
Beta Was this translation helpful? Give feedback.
-
The reason is that the .gitmodules file that contain information about the subdatasets in vanuatu_metrics is somehow pointing to solomon rather than vanuatu. I suggest to |
Beta Was this translation helpful? Give feedback.
-
If needing your nested dataset to refer to a specific branch:
|
Beta Was this translation helpful? Give feedback.
-
In many of our projects, we just want recording-level metrics -- eg how many vocalizations (in total or per hour) the child produced. In this S&T, I explain how I set up one such analysis repository. I'm drawing code from a private repo (LAAC-LSCP/el1000-metrics) -- so credit for that code goes to @lucasgautheron ; my own analysis will be public, but the dataset on which it is based is not (yet) public.
1. Set up a YODA repo
In order to share with others & get feedback, I followed these instructions to publish to GIN, except that:
I got the following messages, which are normal:
Now that means that when I'm ready to share my progress, I do:
Another time I wanted to share this with my colleagues at ACLEW, so I followed instead these instructions to publish to github. I had to generate a token in github here (which allows me to access the github API).
I got these messages:
Note that where it says
token:
the system waited until I copy-pasted my token.2. Import analyses etc from a similar analysis repo
I first made a local copy of the similar analysis repo somewhere else (i.e., not in the repo I just created):
After poking around, I decided to:
And then open in a text editor and make changes to makefile & readme.
3. Nest the dataset
I created folders for input and output:
I'll put it inside input, per YODA recommendations.
4. Run analyses
As described in the README inside the repo you just created, you launch this from the top level of your analysis repo, with:
I got a series of lines like:
Those are just to be ignored...
5. Run reliability/accuracy analyses TBD
6. Embedding these analyses in a paper
For this paper, we are working in a dropbox folder and it is not convenient to migrate to git. So I decide to
embed a clone of these analyses in my dropbox folder. This simply means installing the ANALYSIS datalad dataset;
notice I ALSO use the -r switch because I'm going to use directly the original dataset, specifically to be able
to extract child-level and recording-level metadata.
And done!
Now when I have an update in solis_metrics that I want to take into account, I can cd into it, and do
git pull
. But I can also stick to the current copy, so that my analyses are reproducible.Troubleshooting
Permission denied
It may be that analyses get a permission error when I tried to write a file:
PermissionError: [Errno 13] Permission denied: 'some-filename'
To fix that I did:
datalad unlock output
could not get content
When trying to get files I got:
[WARNING] could not get some content in /Users/acristia/Documents/gitrepos/solis_metrics/input/solomon/annotations/vtc/converted
As documented in #252, this was due to mistakenly installing using https, rather than ssh
the call to child-project didn't work
I doubt others will have this problem, but if you do see:
after your call
child-project metrics input/
etc, that means child-project is outdated and doesn't have the metrics option. Fixed with:Beta Was this translation helpful? Give feedback.
All reactions