execution of code in quarto cell sometimes works, sometimes does not

[rmflight]

Positron Version: 2024.07.0 build 21

Code - OSS Version: 1.90.0 Commit: 82b335e27545e804b7fd806db00477a6eb49951d Date: 2024-07-09T03:06:22.884Z Electron: 29.4.0 Chromium: 122.0.6261.156 Node.js: 20.9.0 V8: 12.2.281.27-electron.0 OS: Linux x64 6.9.3-76060903-generic R: 4.3.0

Steps to reproduce the issue:

1. Create a new quarto document
2. Insert a code cell
3. Try to evaluate code by line or piped statement
4. Only partial evaluation happens

Screencast from 07-09-2024 04:13:08 PM.webm

In this example, I tried to run the code block containing

data(mtcars)

using Ctrl+Enter. But only the data part is called. Another code block, with a library call, seemed to work fine. Also, executing the full block, via Ctrl+Shift+Enter also works (not shown in recording).

I've also had this fail for using cells containing code that spans multiple lines with pipes, and only a single or a couple of lines of the pipe will be executed.

What did you expect to happen?

Expect full code statement to be evaluated.

Were there any error messages in the output or Developer Tools console?

No.

[jmcphers]

I couldn't reproduce this after typing out the example from the screenshot above.

@rmflight Could you share any other examples, maybe one with the partial pipes?

[rmflight]

Sure thing. I also checked the logs, and there is something weird in there.

Better example

---
title: "Untitled"
format: html
---

```{r}
#| label: test-block
data("mtcars")

#| label: run-ggplot
library(ggplot2)

#| label: test-piped
mtcars |>
    dplyr::filter(mpg > 20) |>
    dplyr::summarise(n = dplyr::n())

#| label: another-piped
mtcars |>
    dplyr::filter(mpg < 20) |>
    dplyr::summarise(n = dplyr::n())

Screencast of today: 

[Screencast from 07-10-2024 03:28:48 PM.webm](https://github.com/posit-dev/positron/assets/1509626/01a9ae21-e798-422a-ab01-a908fcf9ac3c)

And log output. 

[log_2024-07-10-15-25.log](https://github.com/user-attachments/files/16168078/log_2024-07-10-15-25.log)

You can see that only one line of the pipe is being sent at a time, when I expect the full thing to be sent.

In addition, ark is restarting my R kernel when it tries to access the variables so I can see them in my dplyr call. 

Also, I'm running R 4.3.0, in case that matters.

[juliasilge]

Might this be related to #3792 as well, and something going wrong with the mapping between .qmd and .R?

[rmflight]

I'm not sure what has changed from build 21 to build 52, but this issue seems to have gone away as of now.

I installed latest release this morning, and it seems to be behaving properly in my .qmd documents now.

[coreysparks]

I am still seeing this issue with today's build, no rendering of output within a quarto document

mean(rnorm(10))
hist(rnorm(100))

should look like this from Rstudio

[juliasilge]

Ah @coreysparks I believe you are expecting to see inline output for a Quarto document, which is not what this issue is about. You can read more about using Quarto in an inline mode (vs. a "send to console" mode) in #1418, #1533, #3718, and #4033.

[rmflight]

Positron Version: 2024.08.0 build 24

I keep downloading the new releases to see if this has been fixed yet, and to see if I can figure out where exactly the issue comes from.

I just noticed today, that if I open a project with a quarto document in it, I can pass any lines already written in the document when I opened the project, but if I try and write any new r code blocks, or start a new document, it doesn't work.

Nothing shown in the output log besides R starting up.

[DavisVaughan]

I'm seeing this in the logs that @rmflight posted, making me think this is indeed related to #3792

[R]   2024-07-10T19:29:37.116795Z ERROR  Requesting line 27 but only 17 lines exist. Document: '#
[R] #
[R] #
[R] #
[R] #
[R] #
[R] #| label: test-block
[R] data("mtcars")
[R] #
[R] #
[R] #
[R] #| label: run-ggplot
[R] library(ggplot2)
[R] #
[R] #
[R] #
[R] '.
[R]     at crates/ark/src/lsp/encoding.rs:84

It would be nice if we saw a full traceback in the logs here, so we can determine where exactly this is coming from in ark. (The error is here but I'd like the full call stack https://github.com/posit-dev/ark/blob/5c803695fca80ef5514db0a7224bd008e003898b/crates/ark/src/lsp/encoding.rs#L84)

So one thing we should definitely do is try and either convert this error log to an actual error, or maybe try and force a backtrace capture right here and log it (I think we can do that), and then request that @rmflight reupload logs after triggering it again.

[rmflight]

I'm not sure if the Positron Version: 2024.09.0 build 27 has what you wanted @DavisVaughan , or if I'm waiting on another build, but I went ahead an forced the error again.

Logs attached from today.

log_2024-09-13-14-56.log

[DavisVaughan]

Thanks, looks like the backtrace itself didn't prove that useful (probably because you are on a release build with some debug symbols stripped out) but this is still useful for us to start playing with

[R]   2024-09-13T18:52:04.611438Z ERROR  Requesting line 33 but only 31 lines exist.
[R] 
[R] Document:
[R] #
[R] #
[R] #
[R] #
[R] #
[R] #
[R] #| label: test-block
[R] data("mtcars")
[R] #
[R] #
[R] #
[R] #| label: run-ggplot
[R] library(ggplot2)
[R] #
[R] #
[R] #
[R] #| label: test-piped
[R] mtcars |>
[R]     dplyr::filter(mpg > 20) |>
[R]     dplyr::summarise(n = dplyr::n())
[R] #
[R] #
[R] #
[R] #| label: another-piped
[R] mtcars |>
[R]     dplyr::filter(mpg < 20) |>
[R]     dplyr::summarise(n = dplyr::n())
[R] #
[R] #
[R] #
[R] 
[R] 
[R] 
[R]     at crates/ark/src/lsp/encoding.rs:86

In particular, reverse engineering the qmd...

---
title: "Untitled"
format: html
---

```{r}
#| label: test-block
data("mtcars")

#| label: run-ggplot
library(ggplot2)

#| label: test-piped
mtcars |>
  dplyr::filter(mpg > 20) |>
  dplyr::summarise(n = dplyr::n())

#| label: another-piped
mtcars |>
  dplyr::filter(mpg < 20) |>
  dplyr::summarise(n = dplyr::n())

I still can't force the panic when playing around with this qmd but maybe someone on the team can

[rmflight]

It seems to evaluate pre-existing code blocks OK when I open the file. It's when I add a new code block that it seems to have trouble with.

Also, I have custom snippet for QMD files, that when I hit "r" -> "tab" I get the option to insert a qmd code block. 😉

[rmflight]

If someone wanted to email me instructions on getting a non-release version installed (my email is listed on my profile) I could totally do that and see if we can get more output that helps things.