Publishing expanded reading history

[handwerkerd]

I realized I didn't run make site-publish before merging.
@tsalo Thoughts on best practices for this? Once someone approves a PR, the submitter then runs make site-publish on the same PR before merging?

[tsalo]

I think that's a good policy.

I'm a little worried about the size of this change. I'd expect it to only affect the built version of the recommended reading page, but it looks like there are a lot of other changed files.

[handwerkerd]

It looks like some files have hashes in their file names so maybe they change?

Another factor is floating point errors. Every execution will generate slightly different volumes, so they are effectively new files (unless just a different time stamp is enough to add a file to the commit.) Is there any non-painful way for make site-publish to not replace png images that are identical to the previous images within some tolerance threshold?

That said, some images do look different. For example:
docs/_images/fig_s0_bold_single-echo.png
docs/_images/fig_t2s_bold_single-echo.png

This is random each iteration, with no set seed, so that's what's happening.

multi-echo-data-analysis/content/Signal_Decay.md

Line 271 in 85d75fb

ts = np.random.normal(loc=0, scale=1, size=(N_VOLS + 20,))

I'd expect anything post ICA to actually change.