How to pass length-1 vector as vector?

[YStrauchP4]

Hi!

R has this weird quirk that everything is a vector, even strings.

If my template uses a for loop on a vector, it succeeds if there's more than one entry but crashes if its length is just one, I guess some internal logic casts it to a string/number instead of an iterable with one element. This makes my templates crash randomly and I had to implement a helper function to manually cast vectors of length one to a list when I need to iterate over it. That's quite ugly...

I'm not sure if you can do anything about this, I guess you need to guess the data types and convert them for the C++ data types and you can't know if someone wants to iterate over a vector or use it as a string, but I thought I leave a ticket as this is an annoying caveat of R. Maybe add some documentation if you can't fix it haha

[davidchall]

Hi @YStrauchP4,

You're absolutely right that a length-1 vector is automatically converted to a scalar to be used by the template engine. We needed to choose a default behavior for this ambiguous scenario, and this was the most common usage pattern. This behavior is documented here.

However, I agree there are many cases where you want to pass a vector. And that vector will sometimes have 1 element! 😄 Fortunately, you can explicitly request this behavior using the I() function (aka the "as-is" operator).

library(jinjar)

template = "{% for person in people -%}
Hello {{ person }}
{%- endfor -%}"

# vector with 2 elements
render(template, people = c("Woody", "Buzz")) |> writeLines()
#> Hello Woody
#> Hello Buzz

# vector with 1 element
render(template, people = c("Andy")) |> writeLines()
#> Error in `render()` at jinjar/R/render.R:37:2:
#> ! Problem encountered while rendering template.
#> Caused by error:
#> ! Object must be an array.
#> ℹ Error occurred on line 1 and column 15.

# explicitly prevent conversion to scalar
render(template, people = I(c("Andy"))) |> writeLines()
#> Hello Andy

^{Created on 2022-12-02 with reprex v2.0.2}

I totally agree this feature needs to be documented though - I'll add this in the next release. Thanks for the report!

[YStrauchP4]

Thanks for your quick response!
Using the as-is operator is certainly less ugly than my helper function, wasn't aware of this operator and will certainly use it instead! Thanks for adding it to the docs!

[33Vito]

Hi David,

Great package! Love it as a long-time Jinja user, really hope more features will become available (e.g. the '|' filter feature in Python)

In this case, I just want to check if there is a way for this to work with a list input loaded from external? In my workflow, I use a YAML file to contain all the parameters required by the Jinja template. The YAML object loaded into R becomes a deeply nested list.

This following will work

render(template, !!!list(people = I(c("Andy")))) |> cat()

But I was looking for a way to define the parameters completely inside YAML so R does not have to mess with it.

[davidchall]

Hi @33Vito - So in your use case, you want to specify that all length-1 vectors should be treated as vectors. Is that right?

[33Vito]

Hi @33Vito - So in your use case, you want to specify that all length-1 vectors should be treated as vectors. Is that right?

Yes

[davidchall]

Hi @33Vito - you can apply I() to every value in your deeply nested list using purrr::modify_tree().

Here's an example, demonstrating that field1 and field3 are both treated as vectors.

library(jinjar)

# data stored in nested list (read from YAML)
data <- list(field1 = "value1", field2 = list(field3 = "value3"))

# wrap every value in I() to preserve vectors
data2 <- purrr::modify_tree(data, leaf = I)

# inspect data structures
str(data)
#> List of 2
#>  $ field1: chr "value1"
#>  $ field2:List of 1
#>   ..$ field3: chr "value3"
str(data2)
#> List of 2
#>  $ field1: 'AsIs' chr "value1"
#>  $ field2:List of 1
#>   ..$ field3: 'AsIs' chr "value3"

# render example template
template <- "
{% for val in field1 %}{{ val }}{% endfor %}
{% for val in field2.field3 %}{{ val }}{% endfor %}
"

render(template, !!!data2)
#> [1] "\nvalue1\nvalue3\n"

^{Created on 2023-03-06 with reprex v2.0.2}

[33Vito]

This is awesome. Thanks, mate.