Is there a way to pass variable between different language code cells (from or to bash)?

[aborruso]

Hi,
is there a way to do something like this?

```{bash}
var="hello"
```

```{python}
print(var)
```

[mcanouil]

Languages are not kernel.
You can only use one kernel in a document, i.e., knitr, jypter, etc.

You need to use the proper syntax of the language in which you want to retrieve the information and/or make sure to export properly what you want to use.
For instance, look at how data are passed to Observable in the documentation: https://quarto.org/docs/computations/ojs.html.

Environment variables can usually be used by any languages, but you can store anything in it as long as it is a character string.

[aborruso]

As illustrated, this is only possible within the same session.

I swear I get it.

For ojs_define is only for R/Python at the moment. There are too many languages, too many object structure to be handle by Quarto which is a language agnostic tool. This being said, nothing forbids anyone to create/extend the reach of ojs_define or do similar job as reticulate and rpy2 to only quote those.

Unfortunately I am not able to do it, otherwise I would write it myself. I took the liberty of repeating myself, only because I think that building bridges to bash is an important thing, because you can do a lot of great data things there.

Thank you @mcanouil

[mcanouil]

I agree that extending to bash (also PowerShell for Windows users) could be great, but not sure it's going to happen at least in a near future for ojs_define.

Note that contrary to ojs_define defined by Quarto, for the bash thing using knitr as engine, the best target for a feature request would be the knitr repository and when using jupyter, well the official repository for it.
Quarto only uses those engine/kernels. 😉

[cscheid]

(I've edited your posts to remove typos, just FYI)

[aborruso]

@mcanouil @cscheid in some way I found what I was looking for, using system in r.

One of my goals is sometimes to use a bash data analysis utility and collect the output in a variable and pass it to r, to use the convenient syntax The sum of this is `r sum` .

In the below example I use Miller to make a sum. I get the result and I use it in markdown.

And using reticulate I can pass it to python.

I'm sure you don't like it and it's not elegant, but since I can't do it from a bash block, I've been looking for something like python magic in a r block.

A stupid question: how can I not put the absolute path of the CSV file, and insert a variable that prints the current path (the qmd and the CSV are in the same folder)?

Best regards

```{r}
sum <- system('mlr --c2n stats1 -a sum -f a /home/andy/quarto_test/input.csv', intern = TRUE)
```

The sum is `r sum`.

```{python}
sumpy = r.sum
print(sumpy)
```

[cscheid]

Something like this?

```{r}
cmd <- paste0(c("mlr --c2n stats1 -a sum -f a ", getwd(), "input.csv"), collapse="")
sum <- system(cmd, intern = TRUE)
```

But I'm a bit confused, because I think that simply omitting the path should work; the working directory for the system() call should be that of the document.