v2 migration feedback

[Keats]

I think I've covered everything in the MIGRATION.md file but let me know if there are things that are missing/unclear in this issue.

Also if something is now impossible to do in v2 that could be done in v1.

[sanpii]

First, thank you for this release 🎉

I easily migrate my projects using tera, but I didn’t use the MIGRATION.md file because it’s very detailed. For me, It was easier to try to compile me project and see what failed.

I would have liked a summary of breaking changes. Something like that:

• strict mode;
• built-in function/filter moves to tera-contrib;
• new function/filter arguments;
• Macros are replaced by components.

I had one trouble to migrate a section of code that used from_value because I can’t impl ArgFromValue for my types.

With tera 1.0:

let pager: Pager = tera::from_value(value.clone())?.unwrap();
let base_url: String = tera::from_value(value.clone())?.unwrap_or_default();

With tera 2.0, I would like to write:

let pager: Pager = kwargs.must_get("pager")?;
let base_url: String = kwargs.get("base_url")?.unwrap_or_default();

I created a new type and deserialized it:

#[derive(serde::Deserialize)]
struct Param {
    pager: Pager,
    #[serde(default)]
    base_url: String,
}

let param: Param = kwargs.deserialize()?;

[Keats]

I would have liked a summary of breaking change

Isn't that this section?

I had one trouble to migrate a section of code that used from_value because I can’t impl ArgFromValue for my types.

Ah I didn't think passing objects as params. We could add a Deser wrapper or a get/must_get_deserialized or something.

[rochacbruno]

Hi, thanks for this release! very interesting features.

I wonder if you plan to keep the v1 branch and backports to v1 releases?

NEVERMIND! I think I found the workaround, I am pre-processing templates to make those compatible rochacbruno/marmite#470

Thanks

---

I maintain a Blog Generator, I have multiple users that uses customized templates, also the shortcodes feature on my system is entirely based on macros.

I want users to migrate gradually to new templates, I prefer subscription indexing foo[1] instead of current foo.1. However this will be a huge effort, as I dont know how many users are having custom templates, fragments and shortcodes.

So, I was thinking about adding improvements that are non-breaking to the v1 as it would give me more time to communicate and plan some tool to help users migrate. (I may create a CLI options that would rewrite templates, or add a template pre-processing tool to the SSG itself)

My other option is to keep a very focused fork and maintain it until I have everyone migrated to 2.0.0 template style.

Thanks for amazing work.

[cs-clarence]

Hi, thanks for the new release!

Is it possible to make the Template struct public again? Previously it as public in v1 and I was using to validate string inputs via parsing:

pub fn validate_str<V: Borrow<str>>(value: &V) -> Result<(), String> {
    let result =
        tera::Template::new("__validate_template__", None, value.borrow());

    if let Err(error) = result {
        return Err(error.to_string());
    }

    Ok(())
}

Also, how what is the recommended approach for creating and extending presets now for v2?
in V1

let mut tera = tera::Tera::default();
tera.extend(FULL_PRESET);

in v2?

let mut tera = FULL_PRESET.clone();

Thanks for the time and effort you spent for this new release.

[Keats]

Is it possible to make the Template struct public again?

Lots of validation happen in the Tera struct when adding templates so I think it's better to just use Tera::default() + add_raw_template(s) than expose Template::new. Template is not even exposed in v2.

Also, how what is the recommended approach for creating and extending presets now for v2?

It kind of depends what you need exactly. You have register_from if you want to only take functions/filters/tests otherwise clone works yes and set_fallback_prefixes to handle the case where some files might be referenced differently from their file path (eg Zola themes).

[GildedHonour]

I think I've covered everything in the MIGRATION.md file

But you haven't.

let templates = Arc::new(tera::Tera::new("web/html_templates/**/*.html").expect("failed to load templates"));

v1.2 - OK
v2.0 -

error[E0061]: this function takes 0 arguments but 1 argument was supplied
  --> src/main.rs:35:30
   |
35 | ...Arc::new(tera::Tera::new("web/html_templates/**/*.html").expect("failed...
   |             ^^^^^^^^^^^^^^^ ------------------------------ unexpected argument of type `&'static str`
   |

Nothing about ::new(...) in https://github.com/Keats/tera/blob/master/MIGRATION.md

[GildedHonour]

@Keats

Lots of validation happen in the Tera struct when adding templates so I think it's better to just use Tera::default() + add_raw_template(s) than expose Template::new.

Now one will have to enumerate every single fиск!ng template in each directory and sub-directory, call a function to add it. Or write a helper that will traverse a template directory. Right?

[GildedHonour]

Found it - glob_fs. Resolved!

[GildedHonour]

Yet, how to a) use it b) migrate
still not described. The feature isn't present in the documentation either; briefly mentioned, once, in the migration guide

[Keats]

It's in https://keats.github.io/tera/#getting-started
It was missing in lib.rs but I've added the same features blurb yesterday I think so it will be present once there is a new release. I thought docs.rs would flag some functions needing a feature, eg https://docs.rs/tera/latest/tera/struct.Tera.html#method.load_from_glob

[asimpletune]

Is there no replacement for get_env?

[Keats]

Not built-in. I might add it to Zola though.

[kengreim]

Does dot notation to access array members by index not work anymore? Something like boxes.0 used to work but now gives an error Found integer but expected identifier." but boxes[0] works. I didn't see anything in the v2 docs or the migration readme about this.

Edit: boxes is Vec<String>, just to be clear.

[Keats]

Ah no boxes.0 will not work anymore. I need to add that in the migration file.