hive-forge: add --since cursor paging to comments and timeline

hive-forge/src/verbs/comments.rs

@@ -1,65 +1,92 @@
 //! `comments <number> [--limit N | --tail N]` — list comments on an
-//! issue or PR. Replaces the curl fallback; `--tail`
-//! handles the paging-for-long-threads
-//! awkwardness).
+//! issue or PR. Replaces the curl fallback.
 //!
-//! - `--limit N` (default 50, Forgejo's cap) returns the first N
-//!   comments — same shape this verb has always had.
-//! - `--tail N` returns the *last* N comments in chronological
-//!   order. Reads the issue's `comments` count first to compute
-//!   which page contains the tail, then fetches only `ceil(N/50) +
-//!   1` pages. No upstream cap — the work is bounded by `N`, not
-//!   by the thread's length, so it stays cheap even on threads with
-//!   thousands of comments. Use this for "what was the conclusion
-//!   on this long thread?" without scrolling through the whole
-//!   history.
+//! - `--limit N` (default 10, was 50) returns the first N comments. A
+//!   big default was a silent trap on long threads: a plain `comments`
+//!   call reads as "I've read this thread" when it's really "I've read
+//!   the oldest N of it".
+//! - `--tail N` returns the *last* N comments, chronological. Reads the
+//!   issue's `comments` count first to compute which page holds the
+//!   tail, then fetches only `ceil(N/50) + 1` pages — bounded by `N`,
+//!   not thread length, so it stays cheap on threads with thousands of
+//!   comments.
+//! - **The count of comments outside whatever window is shown is
+//!   always reported** — never a silent truncation.
+//! - `--since <RFC3339>` filters to comments at or after that
+//!   timestamp instead of a head/tail window — a cursor: feed the
+//!   last-seen row's own `created_at` back in next time to fetch only
+//!   what's new. Mutually exclusive with `--tail`. No total exists for
+//!   a since-filtered query, so this uses the same over-fetch-by-one
+//!   trick `timeline`'s `--limit` does, and `--limit` is clamped the
+//!   same way (see `crate::verbs::MAX_LIMIT`).
 //!
-//! For PRs, review *bodies* (the summary text submitted with an
-//! approve / request-changes / comment review) are merged in too:
-//! they live in the `pulls/<n>/reviews` object, NOT the
-//! issues/comments thread, so plain comment listings used to miss
-//! them entirely and reviewers/authors silently lost feedback —
-//! the gap this fix closes. They're always included regardless of
-//! `--limit`/`--tail`
-//! (reviews are few + high-signal) and tagged `[review: STATE]` so
-//! they're distinguishable from issue-thread comments.
+//! Review *bodies* on PRs are always merged in too (`pulls/<n>/reviews`
+//! isn't the issues/comments thread, so a plain listing used to miss
+//! them), tagged `[review: STATE]`.
 //!
-//! Use the global `--json` flag for JSON output.
+//! `--json` output is an object (`{"comments": [...], "more_before": N,
+//! "more_after": N, "since_more": bool}`), not a bare array, so a
+//! script can read the truncation info too.
 
 use anyhow::Result;
 use clap::Args as ClapArgs;
 use forgejo_api::structs::IssueGetCommentsQuery;
 use serde_json::{Value, json};
+use time::OffsetDateTime;
 
 use crate::client::{Client, index};
 use crate::notify;
-use crate::verbs::{print_json, rfc3339};
-
-/// Forgejo's per-page comment cap. The API caps `limit` at 50 even
-/// if a higher value is requested; pin it explicitly so the math
-/// downstream doesn't depend on a hidden default.
-const PAGE_SIZE: usize = 50;
+use crate::verbs::{MAX_LIMIT, PAGE_SIZE, clamp_limit, parse_rfc3339, print_json, rfc3339};
 
 #[derive(ClapArgs)]
 pub struct Args {
     /// Issue or PR number.
     pub(crate) number: u64,
-    /// Number of comments from the start of the thread (max 50).
-    /// Mutually exclusive with `--tail`.
-    #[arg(long, default_value_t = 50, conflicts_with = "tail")]
+    /// Number of comments from the start of the thread, or (with
+    /// `--since`) the most this call returns — capped at
+    /// [`crate::verbs::MAX_LIMIT`] in the latter case. Mutually
+    /// exclusive with `--tail`.
+    #[arg(long, default_value_t = 10, conflicts_with = "tail")]
     limit: u64,
     /// Return the last `N` comments (chronological). Mutually exclusive
-    /// with `--limit`.
-    #[arg(long)]
+    /// with `--limit`/`--since`.
+    #[arg(long, conflicts_with = "since")]
     tail: Option<usize>,
+    /// Only show comments at or after this RFC3339 timestamp (same
+    /// format this verb's own output prints). Mutually exclusive with
+    /// `--tail`.
+    #[arg(long)]
+    since: Option<String>,
 }
 
 pub fn run(client: &Client, args: Args) -> Result<()> {
     let repo = client.repo();
-    let thread = match args.tail {
-        Some(n) => fetch_tail(client, args.number, n)?,
-        None => fetch_head(client, args.number, args.limit)?,
-    };
+    // Truncation info: what sits outside the window we're about to show.
+    // A `--tail` window has nothing after it (it ends at the thread's
+    // current end); a `--limit`/default window has nothing before it (it
+    // starts at the thread's beginning); a `--since` window is a boolean
+    // "there's more" (no total exists for a since-filtered query) rather
+    // than an exact count. `more_before`/`more_after` are derived from
+    // `thread.len()` rather than the requested `n`/`limit` so a
+    // shrunk-under-us thread (comments deleted mid-fetch) still reports
+    // accurately instead of the number we merely asked for.
+    let (thread, more_before, more_after, since_more, since_clamped) =
+        if let Some(since_str) = &args.since {
+            let since = parse_rfc3339(since_str)?;
+            let (limit, clamped) = clamp_limit(args.limit);
+            let (thread, more) = fetch_since(client, args.number, since, limit)?;
+            (thread, 0, 0, more, clamped)
+        } else if let Some(n) = args.tail {
+            let total = fetch_total(client, args.number)?;
+            let thread = fetch_tail(client, args.number, n, total)?;
+            let more_before = total.saturating_sub(thread.len());
+            (thread, more_before, 0, false, false)
+        } else {
+            let total = fetch_total(client, args.number)?;
+            let thread = fetch_head(client, args.number, args.limit)?;
+            let more_after = total.saturating_sub(thread.len());
+            (thread, 0, more_after, false, false)
+        };
     // Merge in PR review bodies (empty for issues — degrades to a
     // no-op) so review feedback isn't silently dropped.
     let comments = merge_chronological(thread, fetch_review_bodies(client, args.number));
@@ -82,7 +109,13 @@ pub fn run(client: &Client, args: Args) -> Result<()> {
                 })
             })
             .collect();
-        print_json(&Value::Array(trimmed))
+        print_json(&json!({
+            "comments": trimmed,
+            "more_before": more_before,
+            "more_after": more_after,
+            "since_more": since_more,
+            "since_limit_clamped": since_clamped,
+        }))
     } else {
         for c in &comments {
             let user = c
@@ -100,10 +133,58 @@ pub fn run(client: &Client, args: Args) -> Result<()> {
             }
             println!();
         }
+        if since_clamped {
+            println!(
+                "(--limit {} is above the {MAX_LIMIT} cap --since can reliably detect truncation at — clamped)",
+                args.limit
+            );
+        }
+        if since_more {
+            println!(
+                "(more comments since this timestamp not shown — raise --limit or bump --since; exact count not available)"
+            );
+        }
+        if let Some(note) = truncation_note(more_before, more_after) {
+            println!("{note}");
+        }
         Ok(())
     }
 }
 
+/// A one-line truncation note when the shown window doesn't cover the
+/// whole thread — never a silent gap. `None` when both are zero (the
+/// whole thread rode the wire). Returns a `String` rather than printing
+/// directly so the three cases are unit-testable.
+fn truncation_note(more_before: usize, more_after: usize) -> Option<String> {
+    match (more_before > 0, more_after > 0) {
+        (false, false) => None,
+        (true, false) => Some(format!(
+            "({more_before} earlier comment(s) not shown — use --tail to see the start of the thread)"
+        )),
+        (false, true) => Some(format!(
+            "({more_after} more comment(s) not shown — use --tail N to see the latest instead)"
+        )),
+        (true, true) => Some(format!(
+            "({more_before} earlier + {more_after} later comment(s) not shown — use --tail N)"
+        )),
+    }
+}
+
+/// Total comment count on an issue/PR, straight off the issue object.
+/// Used both to plan `--tail`'s pagination and to report how many
+/// comments sit outside whatever window ends up shown.
+fn fetch_total(client: &Client, number: u64) -> Result<usize> {
+    let (owner, name) = client.owner_repo()?;
+    let issue = client
+        .api()
+        .issue_get_issue(owner, name, index(number)?)
+        .send()?;
+    Ok(issue
+        .comments
+        .and_then(|c| usize::try_from(c).ok())
+        .unwrap_or(0))
+}
+
 /// Serialize a typed comment page back to the JSON `Value` shape the
 /// merge + render pipeline works on (the structs serialize to the API
 /// wire shape, so downstream field access is unchanged).
@@ -197,32 +278,26 @@ fn fetch_head(client: &Client, number: u64, limit: u64) -> Result<Vec<Value>> {
 }
 
 /// Fetch the last `n` comments on an issue/PR in chronological order.
+/// `total` (the thread's comment count, from [`fetch_total`]) drives the
+/// pagination plan and is the caller's, not fetched again here.
 ///
 /// Forgejo orders `/issues/<n>/comments` oldest-first and has no
 /// `direction=desc` knob, so naive "page everything and slice" pages
 /// from the WRONG end on long threads — the first 1000 comments
-/// instead of the last `n`. Fix: read the issue's `comments` count
-/// first to know how many exist, then start paginating from the
-/// page that contains item `total - n`. Work is bounded by
-/// `ceil(n/50) + 1` page fetches, regardless of thread length.
-fn fetch_tail(client: &Client, number: u64, n: usize) -> Result<Vec<Value>> {
-    if n == 0 {
+/// instead of the last `n`. Fix: use `total` to know how many exist,
+/// then start paginating from the page that contains item
+/// `total - n`. Work is bounded by `ceil(n/50) + 1` page fetches,
+/// regardless of thread length.
+fn fetch_tail(client: &Client, number: u64, n: usize, total: usize) -> Result<Vec<Value>> {
+    if n == 0 || total == 0 {
         return Ok(Vec::new());
     }
     let (owner, name) = client.owner_repo()?;
     let idx = index(number)?;
-    let issue = client.api().issue_get_issue(owner, name, idx).send()?;
-    let total = issue
-        .comments
-        .and_then(|c| usize::try_from(c).ok())
-        .unwrap_or(0);
-    if total == 0 {
-        return Ok(Vec::new());
-    }
     // Cap `n` at the actual total so the math below stays in range
     // when the caller asks for more comments than exist.
     let n = n.min(total);
-    let page_size = PAGE_SIZE;
+    let page_size = usize::try_from(PAGE_SIZE).unwrap_or(usize::MAX);
     // 0-based index of the first comment we want; integer-divide to
     // get the 1-based page that contains it.
     let start_idx = total - n;
@@ -251,6 +326,36 @@ fn fetch_tail(client: &Client, number: u64, n: usize) -> Result<Vec<Value>> {
     Ok(merged.into_iter().skip(overshoot).collect())
 }
 
+/// Fetch comments at or after `since`, capped at `limit` (already
+/// clamped to [`crate::verbs::MAX_LIMIT`] by the caller). No total exists
+/// for a since-filtered query — Forgejo's API reports none — so this
+/// uses the same over-fetch-by-one trick `timeline`'s `--limit` does:
+/// ask for `limit + 1`, and a full extra row means there's more. Returns
+/// `(comments, more)`, `more` a boolean rather than an exact count.
+fn fetch_since(
+    client: &Client,
+    number: u64,
+    since: OffsetDateTime,
+    limit: u64,
+) -> Result<(Vec<Value>, bool)> {
+    let (owner, name) = client.owner_repo()?;
+    let query = IssueGetCommentsQuery {
+        since: Some(since),
+        ..Default::default()
+    };
+    let fetch_limit = limit.saturating_add(1);
+    let (_, comments) = client
+        .api()
+        .issue_get_comments(owner, name, index(number)?, query)
+        .page_size(u32::try_from(fetch_limit).unwrap_or(u32::MAX))
+        .send()?;
+    let mut values = to_values(comments)?;
+    let limit = usize::try_from(limit).unwrap_or(usize::MAX);
+    let more = values.len() > limit;
+    values.truncate(limit);
+    Ok((values, more))
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -262,7 +367,7 @@ mod tests {
     /// easy to off-by-one — without touching the network.
     fn tail_plan(total: usize, n: usize) -> (usize, usize) {
         let n = n.min(total);
-        let page_size = PAGE_SIZE;
+        let page_size = usize::try_from(PAGE_SIZE).unwrap_or(usize::MAX);
         let start_idx = total - n;
         let start_page = (start_idx / page_size) + 1;
         let last_page = (total - 1) / page_size + 1;
@@ -330,6 +435,31 @@ mod tests {
         assert_eq!(bodies, vec!["c1", "r1", "c2"]);
     }
 
+    #[test]
+    fn truncation_note_silent_when_nothing_hidden() {
+        assert_eq!(truncation_note(0, 0), None);
+    }
+
+    #[test]
+    fn truncation_note_after_only_points_at_tail() {
+        let note = truncation_note(0, 40).unwrap();
+        assert!(note.contains("40"), "{note}");
+        assert!(note.contains("--tail"), "{note}");
+    }
+
+    #[test]
+    fn truncation_note_before_only_mentions_earlier() {
+        let note = truncation_note(12, 0).unwrap();
+        assert!(note.contains("12"), "{note}");
+        assert!(note.contains("earlier"), "{note}");
+    }
+
+    #[test]
+    fn truncation_note_both_sides_mentions_both_counts() {
+        let note = truncation_note(3, 7).unwrap();
+        assert!(note.contains('3') && note.contains('7'), "{note}");
+    }
+
     #[test]
     fn merge_with_no_reviews_is_identity() {
         // Issues have no reviews → fetch_review_bodies returns empty →

hive-forge/src/verbs/mod.rs

@@ -69,6 +69,56 @@ pub(crate) fn rfc3339(ts: Option<OffsetDateTime>) -> Option<String> {
     ts.and_then(|t| t.format(&Rfc3339).ok())
 }
 
+/// Parse a `--since`/`--before` CLI argument as RFC 3339 — the inverse of
+/// [`rfc3339`], so a value copied straight from this tool's own output
+/// (every row prints its `created_at` in this exact shape) round-trips
+/// without reformatting. A bad value gets a message naming what was
+/// typed, not a bare parser error.
+pub(crate) fn parse_rfc3339(s: &str) -> Result<OffsetDateTime> {
+    OffsetDateTime::parse(s, &Rfc3339)
+        .map_err(|e| anyhow::anyhow!("`{s}` isn't a valid RFC 3339 timestamp: {e}"))
+}
+
+/// Forgejo's per-page cap, shared by every listing verb that over-fetches
+/// by one to detect truncation without an exact total (`timeline`'s
+/// `--limit`, `comments`' `--since`). The API silently clamps a requested
+/// page size to this value, so it's pinned explicitly rather than left as
+/// a hidden default downstream math could drift out of sync with.
+pub(crate) const PAGE_SIZE: u64 = 50;
+
+/// The highest `--limit` an over-fetch-by-one truncation check
+/// (`fetch_limit = limit + 1`) can still detect: `PAGE_SIZE - 1`. At
+/// `limit == PAGE_SIZE` the `+1` request silently clamps to `PAGE_SIZE`
+/// server-side and the truncation check goes blind exactly when there's
+/// the most data to miss.
+pub(crate) const MAX_LIMIT: u64 = PAGE_SIZE - 1;
+
+/// Cap `requested` at [`MAX_LIMIT`], reporting whether it had to. Pure so
+/// the boundary math is unit-testable without a network call.
+pub(crate) fn clamp_limit(requested: u64) -> (u64, bool) {
+    let limit = requested.min(MAX_LIMIT);
+    (limit, limit < requested)
+}
+
+#[cfg(test)]
+mod page_limit_tests {
+    use super::{MAX_LIMIT, PAGE_SIZE, clamp_limit};
+
+    #[test]
+    fn clamp_limit_passes_small_requests_through() {
+        assert_eq!(clamp_limit(10), (10, false));
+        assert_eq!(clamp_limit(MAX_LIMIT), (MAX_LIMIT, false));
+    }
+
+    #[test]
+    fn clamp_limit_caps_requests_above_the_boundary() {
+        // Regression: `limit + 1` must never exceed Forgejo's PAGE_SIZE,
+        // or the over-fetch-by-one truncation check goes silently blind.
+        assert_eq!(clamp_limit(PAGE_SIZE), (MAX_LIMIT, true));
+        assert_eq!(clamp_limit(1000), (MAX_LIMIT, true));
+    }
+}
+
 /// Issue-vs-PR kind, for the `pr <verb>` / `issue <verb>` sub-command
 /// validation.
 #[derive(Clone, Copy)]

hive-forge/src/verbs/timeline.rs

@@ -1,8 +1,8 @@
 //! `timeline <number> [--limit N]` — list timeline events on an
-//! issue or PR. Fills the gap where agents kept
-//! falling back to curl for "who closed this?" / "when was this
-//! labelled?" archaeology. Composes naturally with `view <n>` /
-//! `comments <n>` — separate verb keeps the existing shapes stable.
+//! issue or PR. Fills the gap where agents kept falling back to curl
+//! for "who closed this?" / "when was this labelled?" archaeology.
+//! Composes naturally with `view <n>` / `comments <n>` — separate verb
+//! keeps the existing shapes stable.
 //!
 //! Forgejo's `/issues/{n}/timeline` endpoint returns BOTH the actual
 //! comments AND the event entries (label, assignee, close, reopen,
@@ -10,46 +10,73 @@
 //! a human-readable form by default; pass the global `--json` flag
 //! for the raw API shape.
 //!
-//! `--tail N` is a follow-up (the timeline endpoint doesn't expose a
-//! total-count field so we can't use the count-then-page trick that
-//! `comments --tail` uses; future shape probably mirrors
-//! `comments --tail` once Forgejo grows a `count` query or we accept
-//! the trailing-slice cost).
+//! Default `--limit` is 10 (was 50), same silent-truncation-trap fix
+//! as `comments`. **`more` is a boolean, not an exact count**: unlike
+//! `comments` (which reads the issue's own `comments` field),
+//! the timeline endpoint has no total, and `forgejo-api`'s typed
+//! `.send()` never surfaces response headers to check `X-Total-Count`
+//! against — so this over-fetches by one (`limit + 1`) and reports
+//! "there's more" without saying how much.
+//!
+//! `--limit` is capped at [`crate::verbs::MAX_LIMIT`]: the
+//! over-fetch-by-one trick needs `limit + 1` to fit inside Forgejo's
+//! hard per-page cap, or the response silently clamps and the
+//! truncation check can never fire even when there genuinely is more.
+//!
+//! `--since <RFC3339>` filters to events at or after that timestamp —
+//! a cursor: feed the last-seen row's own `created_at` back in next
+//! time instead of re-running with a higher `--limit`.
 
 use anyhow::Result;
 use clap::Args as ClapArgs;
 use forgejo_api::structs::IssueGetCommentsAndTimelineQuery;
-use serde_json::Value;
+use serde_json::{Value, json};
 
 use crate::client::{Client, index};
-use crate::verbs::print_json;
+use crate::verbs::{MAX_LIMIT, clamp_limit, parse_rfc3339, print_json};
 
 #[derive(ClapArgs)]
 pub struct Args {
     /// Issue or PR number.
     pub(crate) number: u64,
-    /// Page size (Forgejo caps at 50). Returns the first `N` events.
-    #[arg(long, default_value_t = 50)]
+    /// Return the first `N` events, capped at 49 (see the module doc
+    /// comment for why). Default kept small on purpose.
+    #[arg(long, default_value_t = 10)]
     limit: u64,
+    /// Only show events at or after this RFC3339 timestamp (same format
+    /// this verb's own output prints) — pass back the last-seen row's
+    /// `created_at` to fetch only what's new.
+    #[arg(long)]
+    since: Option<String>,
 }
 
 pub fn run(client: &Client, args: Args) -> Result<()> {
     let (owner, name) = client.owner_repo()?;
+    // Clamp before the over-fetch-by-one math below: `limit + 1` has to
+    // stay within Forgejo's per-page cap or the truncation check goes
+    // silently blind right at the boundary — see MAX_LIMIT's doc comment.
+    let (limit, clamped) = clamp_limit(args.limit);
+    let since = args.since.as_deref().map(parse_rfc3339).transpose()?;
+    let query = IssueGetCommentsAndTimelineQuery {
+        since,
+        before: None,
+    };
+    // Over-fetch by one to detect truncation without an exact total —
+    // see the module doc comment for why there's no count query here.
+    let fetch_limit = limit.saturating_add(1);
     let (_, events) = client
         .api()
-        .issue_get_comments_and_timeline(
-            owner,
-            name,
-            index(args.number)?,
-            IssueGetCommentsAndTimelineQuery::default(),
-        )
-        .page_size(u32::try_from(args.limit).unwrap_or(u32::MAX))
+        .issue_get_comments_and_timeline(owner, name, index(args.number)?, query)
+        .page_size(u32::try_from(fetch_limit).unwrap_or(u32::MAX))
         .send()?;
+    let limit = usize::try_from(limit).unwrap_or(usize::MAX);
+    let more = events.len() > limit;
+    let events: Vec<_> = events.into_iter().take(limit).collect();
     // Serialize back to the API's JSON shape so the per-type render
     // arms (and their tests) keep working on plain `Value`s.
-    let v = serde_json::to_value(events)?;
+    let v = serde_json::to_value(&events)?;
     if client.json_mode() {
-        return print_json(&v);
+        return print_json(&json!({ "events": v, "more": more, "limit_clamped": clamped }));
     }
     let Some(events) = v.as_array() else {
         return print_json(&v);
@@ -57,6 +84,15 @@ pub fn run(client: &Client, args: Args) -> Result<()> {
     for ev in events {
         print_event(ev);
     }
+    if clamped {
+        println!(
+            "(--limit {} is above the {MAX_LIMIT} cap this verb can reliably detect truncation at — clamped)",
+            args.limit
+        );
+    }
+    if more {
+        println!("(more activity not shown — raise --limit to see it; exact count not available)");
+    }
     Ok(())
 }