document feed() pipeline

Cargo.lock

@@ -616,7 +616,7 @@ dependencies = [
 
 [[package]]
 name = "markov"
-version = "0.7.3"
+version = "0.7.4"
 dependencies = [
  "async-compression",
  "bzip2-sys",

src/feed.rs

@@ -5,7 +5,26 @@ use sanitise::Sentance;
 
 const FEED_BOUNDS: std::ops::RangeFrom<usize> = 2..; //TODO: Add to config somehow
 
-
+/// Feed `what` into `chain`, at least `bounds` tokens.
+///
+/// # Tokenising
+/// How the tokens are split within this function that operates on single buffers is determined largely by the features `split-sentance` and `feed-sentance` determining the use of the sentance API.
+///
+/// ## Pipeline
+/// Since this is called on single buffers, it happens after the `split-newlines` tokenising if it's enabled, and thus the sentance API is only able to operate on each seperate line if that feature is enabled, regardless of `always-aggre`, or `feed-sentance` or `split-sentance`.
+/// This is the pipeline for just within this function, after splitting through newlines if enabled.
+///
+/// * `feed-sentance`
+/// ** Feed the buffer through the sentance split tokeniser
+/// ** Feed the sentances through the word split tokeniser
+/// ** Feed each collection of words into the chain seperately
+/// * `split-sentance`
+/// ** Feed the buffer through the sentance split tokeniser
+/// ** Feed the sentances through the word split tokeniser
+/// ** Feed the flattened collection into the chain once, concatenated.
+/// * Neither
+/// ** Feed the buffer through the word split tokeniser
+/// ** Feed the collection into the chain
 pub fn feed(chain: &mut Chain<String>, what: impl AsRef<str>, bounds: impl std::ops::RangeBounds<usize>)
 {
     cfg_if! {
@@ -84,7 +103,6 @@ pub async fn full(who: &IpAddr, state: State, body: impl Unpin + Stream<Item = R
 		if #[cfg(feature="split-newlines")] {
 		    for buffer in buffer.split('\n').filter(|line| !line.trim().is_empty()) {
 			feed!(&mut chain, buffer, FEED_BOUNDS);
-			
 		    }
 		} else {
 		    feed!(&mut chain, buffer, FEED_BOUNDS);