MCMC sampling messages not displayed in JupyterLab

Summary:

MCMC sampling progress messages are not displayed in JupyterLab or Jupyter Notebook.
Although the issue is actually caused by Jupyter, redirecting the stdout messages of Stan to the R streams could be a workaround solution.

Description:

Currently Jupyter does not capture OS-level output, this problem is not fixed for 2 years (see IRkernel/IRkernel#416 and jupyter/enhancement-proposals#14). As a consequence, the MCMC sampling messages of Stan are not shown in JupyterLab or Jupyter Notebook.

It is suggested that redirecting the stdout messages to the R streams could solve the problem (see IRkernel/IRkernel#584).

Update:

Sampling messages are displayed when mc.cores = 1 .

Reproducible Steps:

Run any model

Current Output:

No output

Expected Output:

Something like this:

SAMPLING FOR MODEL 'x' NOW (CHAIN 1).

Iteration:     1 / 50500 [  0%]  (Warmup)
Iteration:   501 / 50500 [  0%]  (Sampling)
Iteration:  5550 / 50500 [ 10%]  (Sampling)
Iteration: 10600 / 50500 [ 20%]  (Sampling)
Iteration: 15650 / 50500 [ 30%]  (Sampling)
Iteration: 20700 / 50500 [ 40%]  (Sampling)
Iteration: 25750 / 50500 [ 50%]  (Sampling)
Iteration: 30800 / 50500 [ 60%]  (Sampling)
Iteration: 35850 / 50500 [ 70%]  (Sampling)
Iteration: 40900 / 50500 [ 80%]  (Sampling)
Iteration: 45950 / 50500 [ 90%]  (Sampling)
Iteration: 50500 / 50500 [100%]  (Sampling)

 Elapsed Time: 0.00354 seconds (Warm-up)
               0.273752 seconds (Sampling)
               0.277292 seconds (Total)

RStan Version:

2.17.3

R Version:

R version 3.5.1 (2018-07-02)

Operating System:

Ubuntu 18.04.1

Start by reproducing any model in JupyterLab or Jupyter Notebook with the default parallel sampling, then compare it with mc.cores = 1, where messages are displayed. Review the referenced IRkernel#416, IRkernel#584, and Jupyter enhancement proposal to understand the stdout and R-stream handling; done means sampling progress and elapsed-time messages appear in the notebook.