Static branching exercises error

[sctyner]

Prework

• Read and abide by learndrake's code of conduct.
• Search for duplicates among the existing issues, both open and closed.

Description

The exercises page at https://colorado.rstudio.com/rsc/learndrake/static/ has an error. (4-static/4-static.Rmd)

Reproducible example

Not applicable

Session info

I got the error on Microsoft Edge Version 86.0.622.38 and Google Chrome Version 86.0.4240.75. Running Windows 10 Enterprise.

Expected output

A functioning learnr tutorial

[wlandau]

Thanks for catching that. Not sure what is wrong yet, but for now there is a working instance at https://wlandau.shinyapps.io/learndrakestatic/. Currently checking if I just need to regenerate manifest.json.

By the way, targets is the long-term successor to drake, and it is in the initial stages of rOpenSci review. It solves several of drake's permanent problems, including the reliance on a complicated DSL to declare each target. Static branching in targets is easier to learn and use, more consistent, and more reliable. The actual implementation is a bunch of functions in tarchetypes, notably tar_map() and tar_combine().

That said, drake is never going away, and sticking with it is still a reasonable choice. More details are at https://wlandau.github.io/targets/articles/need.html#drake. But if you do decide to transition from drake to targets, these tables can help.

[wlandau]

@colearendt, do you happen to know what would explain this parsing error? manifest.json appears valid when I parse it locally with jsonlite::fromJSON().

[colearendt]

@wlandau It looks like the issue is here:

learndrake/inst/tutorials/learndrakestatic/manifest.json

Line 1724 in 1d7bd19

"lifecycle": {

(I searched for that version number and then poked around that point in the file). Some libraries (i.e. jsonlite::fromJSON()) are a bit more forgiving on JSON parsing, but may just truncate part of the file. There may be a way to make it more strict though - I'm not sure off-hand.

Was this manifest produced directly by rsconnect::writeManifest()?

(Also, do you have a workshop coming up?)

EDIT: Merp. Never mind. This is a map, not an array 😑 Too early haha. I'm actually suspicious this error is coming from elsewhere.

[wlandau]

Yes, the manifest was generated by rsconnect::writeManifest() (with rsconnect 0.8.16). Removing the lifecycle package entry was worth a shot but does not seem to have worked.

I no longer have drake workshops coming up, but I did teach https://github.com/wlandau/targets-tutorial last week at R/Pharma. The RStudio Cloud space worked super well, and I plan to reuse it for subsequent iterations. the targets workshop is entirely notebook-based and does not have apps.

[colearendt]

Awesome! Good to hear! I believe this may be related:

curl: HTTP 404 https://colorado.rstudio.com/rspm/all/__linux__/bionic/latest/src/contrib/Archive/storr/storr_1.2.4.tar.gz
10/13 11:50:44.438 (GMT)
curl: (22) The requested URL returned error: 404
10/13 11:50:45.441 (GMT)
FAILED
10/13 11:50:45.442 (GMT)
Error in getSourceForPkgRecord(pkgRecord, srcDir(project), availablePackagesSource(repos = repos), : Failed to retrieve package sources for storr 1.2.4 from CRAN (internet connectivity issue?) [1.2.1 is current]
10/13 11:50:45.442 (GMT)
10/13 11:50:45.442 (GMT)
Unable to fully restore the R packages associated with this deployment.
10/13 11:50:45.442 (GMT)
Please review the preceding messages to determine which package
10/13 11:50:45.442 (GMT)
encountered installation difficulty and the cause of the failure.
10/13 11:50:45.452 (GMT)
Warning message:
10/13 11:50:45.452 (GMT)
In getSourceForPkgRecord(pkgRecord, srcDir(project), availablePackagesSource(repos = repos), :
10/13 11:50:45.452 (GMT)
Package version '1.2.4' is newer than the latest version reported by CRAN ('1.2.1') -- packrat may be unable to retrieve package sources.

The packrat restore actually failed for your most recent deployment, because you are using a newer version of the storr package than is available on RSPM. If it's possible to downgrade storr, that might fix the deployment issue. Otherwise, RSPM will likely update storr in a day or so and we can redeploy. It's curious that the roll-back failed with this strange issue.

EDIT: Same for drake too 🙈 An RSPM snapshot should be coming soon.

Error in getSourceForPkgRecord(pkgRecord, srcDir(project), availablePackagesSource(repos = repos), : Failed to retrieve package sources for drake 7.12.6 from CRAN (internet connectivity issue?) [7.12.5 is current]

EDIT2: I confirmed that this error is an error from 2 months ago 🙈 I bumped one version back and it is working. When a new deployment succeeds, it should take precedence. These package versions are preventing new deployments.

39705 has the error

[wlandau]

Thanks, Cole! Looks like it's working now.

[colearendt]

Just to be clear, it's running on the old version right now. I'll make a note to try redeploying the latest tomorrow when RSPM is updated.