Python environment persistance across chunks

[talegari]

Each python chunk starts a own session? Would it be a better idea to maintain a single python session while knitting?

Here is the output of the minimal reproducible result: repr.pdf

### Persistance across chunks with R and Python

Lets define x in R

```{r}
# running R
x <- 1
```

Does R remember?

```{r}
# running R
x
```

yes it does!

```{python}
# running python
y = 1
```

Does python remember?

```{python}
# running python
try:
  print(y)
except:
  print("I do not remember 'y'")
```

No, it does not.

----

```{r}
sessionInfo()
```

```{python}
import IPython
print(IPython.sys_info())
```

[yihui]

This is a known issue and we will definitely improve it (hopefully by the end of this year).

[pjcrosbie]

fwiw, imho, this would mean many using python could switch from juypter to knitr, starting with me. many thanks

[yihui]

We have started working on it: rstudio/reticulate#107

[yihui]

Please test the dev version of reticulate. After devtools::install_github('rstudio/reticulate'), you can set up the python engine via this in your first R code chunk:

knitr::knit_engines$set(python = reticulate::eng_python)

Then later Python code chunks should be executed in the same Python session.

[pjcrosbie]

this installed but did not persist the Python session across chunks in an Rmd file on Windows x64

{r} code chunk

devtools::install_github('rstudio/reticulate')
knitr::knit_engines$set(python = reticulate::eng_python)

Skipping install of 'reticulate' from a github remote, the SHA1 (3f3c2277) has not changed since last install. Useforce = TRUEto force installation
trying:

{python} chunk

x = 99
print(x)

99

{python} chunk

print(x)

Traceback (most recent call last): File "C:\Users\pjc\AppData\Local\Temp\Rtmp46YyUn\chunk-code-f4c63c523eb.txt", line 1, in <module> print(x) NameError: name 'x' is not defined

[yihui]

I cannot reproduce your issue on macOS:

```{r}
knitr::knit_engines$set(python = reticulate::eng_python)
```

```{python}
x = 99
print(x)
```

```{python}
print(x)
```

How exactly did you compile your Rmd document?

[pjcrosbie]

my bad.

1. compiles OK when knited, see eg-knitr-python-01.pdf
2. my reported error is when evaluate chunks one at a time in Rmd doc in RStudio, or run preview of .Rmd in RStudio

Q. is (2) a targeted feature, i.e. persisting across chunks in .Rmd when evaluated individually or previewed in RStudio?

eg-knitr-python-01.pdf

[yihui]

For now, we only support the Knit button to compile the whole report. If you want to run the code chunks one by one in the RStudio IDE, I guess we have not added the support yet. You may file a feature request https://github.com/rstudio/rstudio/issues. Thanks!

[pjcrosbie]

Thank you! Another great job. Appreciate having it in knitr; have filed request for RStudio.

[yihui]

I didn't contribute anything this time. @kevinushey was our hero 😁

[kevinushey]

Support for this feature in RStudio with R Markdown documents / R Notebooks is forthcoming, but we wanted to get this ready for users ASAP regardless!

[yihui]

The python engine in the reticulate package has become the default in the current development version of knitr. I'd truly appreciate it if you can test it before I release a new version of knitr to CRAN.

devtools::install_github("yihui/knitr")

[jamieRowen]

I know this has been marked closed, but I get a similar persistence issue when using child option in chunks. The child document.Rmd will knit by itself, but get object not defined errors when knitting the parent.

Parent contains

```{python child='chapter4.Rmd'}

The confusing part is that this does not seem to be consistently true. For example the specific error I get refers to a chunk which has a reference to an object defined 6 chunks earlier. But the previous 5 chunk also refer to the same object, but give no such issue.

I have the dev version of knitr, rmarkdown and reticulate from github. Any ideas? Is this issue connected to the one closed here? Or is this a separate issue that should be opened?

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue by following the issue guide (https://yihui.org/issue/), and link to this old issue if necessary.