Skip to content

Latest commit

 

History

History
140 lines (112 loc) · 5.43 KB

README.md

File metadata and controls

140 lines (112 loc) · 5.43 KB

badpod

A Rust crate for working with imperfect feeds of podcasts.

This crate should not be used for
❌ working with proper, processed podcast feeds
❌ interfacing with your database

This crate can be used for
✅ interpreting external feeds, often from unknown sources
✅ providing feedback about the contents of the feed

Motivation

On a backend server, which has to communicate with a database, strict schemas are typically used. Therefore, if content from an external source that does not conform to the schema is loaded, one may fail to deserialize that content successfully.

This scenario is very common in podcasting space, where RSS feeds of podcasts may

  • combine multiple standards (namespaces)
  • have some elements missing
  • have some values in wrong data types
  • etc.

In that case, we may want a more flexible, intermediate schema: a schema that does not throw an error the instant it encounters an unexpected value but rather one that deserializes the content that it is able to and stores the content that it failed to deserialize for further cleanup or analysis.

Usage

Including in a project

cargo add badpod

This will include the latest version of the crate in your Cargo.toml file.

Deserializing

let rss = match badpod::from_str(feed_str) {
    Ok(rss) => rss,
    Err(_) => panic!("Something went terribly wrong."),
};

In theory, badpod::from_str should only return an error in two cases:

  • the feed is not valid XML
  • the root element of the feed is not <rss>

Features

Check for presence of tags and attributes

In badpod, every field representing an XML tag is a Vector, and every field representing an XML attribute is an Option. This is to reflect that in XML, tags can be repeated, while attributes---cannot. We don't enforce any requirements on what or how many tags should be in the feed---that's a decision for you to make! Leaving it more flexible (instead of throwing an instant error) also allows providing users with better feedback.

match (value_time_split.remote_item.len(), value_time_split.value_recipient.len()) {
    (0, 0) => println!("Either a single `<podcast:remoteItem>` element or one or more `<podcast:valueRecipient>` elements are required."),
    (1, 0) => println!("You are referencing a remote item! Awesome!"),
    (_, 0) => println!("Only a single `<podcast:remoteItem>` element is allowed."),
    (0, _) => println!("You are sharing value with others during this segment! Nice!"),
    (_, _) => println!("Either a single `<podcast:remoteItem>` element or one or more `<podcast:valueRecipient>` elements can be included."),
};

Note: all fields representing tags are named in the singular form, even though they are vectors.

Deserializing complicated tags

badpod converts complex data in text format to something that is easier to work with. If that is not possible, we provide enums with variant Other, which is meant to represent data that could not be deserialized and the reason for that failed deserialization.

match geo {
    podcast::Geo::Ok {
        // f64
        latitude,
        // f64
        longitude,
        // Option<f64>
        altitude,
        // Option<f64>
        uncertainty,
    } => {
        println!("Successfully extracted geographical coordinates!")
    }
    podcast::Geo::Other((s, reason)) => {
        println!("Could not parse coordinates from \"{s}\": {reason}.")
    }
};

Other variant can be especially useful in enums with many variants. If you don't get a match with Other, you know that deserialization yielded something that is reasonably expected.

match language {
    Language::English(region) => println!("A variant of English!"),
    Language::Lithuanian => println!("Lithuanian!"),
    Language::Other((s, _)) => println!("Unexpected language code \"{s}\"."),
    _ => println!("Some other valid language!"),
};

But just because you match a variant that is not Other does not mean that it is a valid value for you. For example, MimeEnclosure has AudioOpus as one of the variants, but if you require that feeds only contain media files with formats supported by Apple Podcasts, then you will want to reject this. Again, badpod is only a tool for analyzing feeds; you decide what a "proper" feed must look like.

Tag-aware deserialization

Many tags use the same data types but encode them differently. For example, both <podcast:locked> and <itunes:explicit> are essentially boolean values, but the former serializes to "yes"/"no" and the latter to "true"/"false". In badpod, both are deserialized to Bool.

match channel.itunes_explicit.get(0) {
    Some(is_explicit) => {
        match is_explicit {
            Bool::Ok(b) => {
                println!("is explicit? \"{b}\"")
            }
            Bool::Other((s, reason)) => println!("could not parse \"{s}\": {reason}"),
        }
    }
    None => println!("<itunes:explicit> not found."),
};

Printing enums

Although this crate is mainly for deserialization, there are cases when enums need to be converted back to strings. In badpod, all enums have Display trait implemented:

// Outputs "cs".
println!("{}", Language::Czech);

// Outputs "en".
println!("{}", Language::English(LanguageEnglish::Default));

// Outputs "en-gb".
println!("{}", Language::English(LanguageEnglish::UnitedKingdom));