feat(rust): dlp --since — filter URLs by message timestamp

Rust-only extension (no python counterpart): --since 5d / 12h / 2026-10-04
keeps only URLs posted at/after the cutoff. URLs without a timestamp
(e.g. --urls-file lists) are kept with a warning.

version-rust/src/orchestrate.rs

@@ -12,6 +12,7 @@ use std::thread;
 use std::time::Duration;
 
 use anyhow::{Context as _, anyhow};
+use chrono::{DateTime, NaiveDate, NaiveDateTime, Utc};
 use serde_json::Value;
 use sha2::{Digest, Sha256};
 
@@ -36,6 +37,9 @@ pub struct RunOpts {
     pub limit: Option<usize>,
     pub shuffle: bool,
     pub only: Option<String>,
+    /// only URLs posted at/after this point (Rust-only extension: "5d",
+    /// "12h", "2026-10-04", or an RFC 3339 timestamp)
+    pub since: Option<String>,
     pub dry_run: bool,
     pub fresh: bool,
 }
@@ -53,12 +57,56 @@ impl Default for RunOpts {
             limit: None,
             shuffle: false,
             only: None,
+            since: None,
             dry_run: false,
             fresh: false,
         }
     }
 }
 
+/// Parse a `--since` value: relative ("5d", "12h", "30m") or absolute
+/// (date, naive datetime, or RFC 3339). Naive times are read as UTC.
+pub fn parse_since(s: &str) -> anyhow::Result<DateTime<Utc>> {
+    let t = s.trim();
+    for (suffix, dur) in [
+        ("d", chrono::Duration::hours(24)),
+        ("h", chrono::Duration::hours(1)),
+        ("m", chrono::Duration::minutes(1)),
+    ] {
+        if let Some(num) = t.strip_suffix(suffix) {
+            let n: i32 = num.parse().map_err(|_| anyhow!("bad --since value: {s}"))?;
+            return Ok(Utc::now() - dur * n);
+        }
+    }
+    if let Ok(dt) = DateTime::parse_from_rfc3339(t) {
+        return Ok(dt.with_timezone(&Utc));
+    }
+    if let Ok(dt) = NaiveDateTime::parse_from_str(t, "%Y-%m-%dT%H:%M:%S%.f") {
+        return Ok(dt.and_utc());
+    }
+    if let Ok(dt) = NaiveDateTime::parse_from_str(t, "%Y-%m-%d %H:%M:%S") {
+        return Ok(dt.and_utc());
+    }
+    if let Ok(d) = NaiveDate::parse_from_str(t, "%Y-%m-%d") {
+        return Ok(d.and_hms_opt(0, 0, 0).unwrap().and_utc());
+    }
+    anyhow::bail!("bad --since value: {s} (use e.g. 5d, 12h, 2026-10-04, or RFC 3339)")
+}
+
+/// Parse a message timestamp from the export; None -> keep-with-warning.
+pub fn parse_message_ts(ts: &str) -> Option<DateTime<Utc>> {
+    if let Ok(dt) = DateTime::parse_from_rfc3339(ts) {
+        return Some(dt.with_timezone(&Utc));
+    }
+    if let Ok(dt) = NaiveDateTime::parse_from_str(ts, "%Y-%m-%dT%H:%M:%S%.f") {
+        return Some(dt.and_utc());
+    }
+    if let Ok(d) = NaiveDate::parse_from_str(ts, "%Y-%m-%d") {
+        return Some(d.and_hms_opt(0, 0, 0).unwrap().and_utc());
+    }
+    None
+}
+
 /// sha256 of a file, streamed (python sha256_of).
 pub fn sha256_of(path: &Path) -> anyhow::Result<String> {
     let mut file =
@@ -327,6 +375,37 @@ pub fn process_channel(
         .filter(|f| opts.only.as_ref().is_none_or(|o| o.as_str() == f.host))
         .collect();
 
+    let found: Vec<FoundUrl> = if let Some(since) = &opts.since {
+        let cutoff = parse_since(since)?;
+        let before = found.len();
+        let kept: Vec<FoundUrl> = found
+            .into_iter()
+            .filter(|f| match &f.timestamp {
+                Some(ts) => match parse_message_ts(ts) {
+                    Some(dt) => dt >= cutoff,
+                    None => {
+                        log::warn!("Unparseable timestamp {ts:?} for {} - keeping it", f.url);
+                        true
+                    }
+                },
+                None => {
+                    // e.g. --urls-file lists carry no attribution
+                    log::warn!("No timestamp for {} - keeping it", f.url);
+                    true
+                }
+            })
+            .collect();
+        log::info!(
+            "--since {}: kept {} of {} URLs (rest are older)",
+            since,
+            kept.len(),
+            before
+        );
+        kept
+    } else {
+        found
+    };
+
     let mut gofile_urls: Vec<FoundUrl> = Vec::new();
     let mut z0807_urls: Vec<FoundUrl> = Vec::new();
     for f in &found {
@@ -1025,4 +1104,38 @@ mod tests {
         fs::write(&p, r#"[{"channel_id": "2"}]"#).unwrap();
         assert!(load_subscriptions(&p).is_err());
     }
+
+    #[test]
+    fn parse_since_relative_and_absolute() {
+        let now = Utc::now();
+        let cutoff = parse_since("5d").unwrap();
+        assert!(cutoff <= now);
+        assert!(cutoff > now - chrono::Duration::days(6));
+        assert!(parse_since("12h").unwrap() > now - chrono::Duration::hours(13));
+        assert_eq!(
+            parse_since("2026-10-04").unwrap().to_rfc3339(),
+            "2026-10-04T00:00:00+00:00"
+        );
+        assert_eq!(
+            parse_since("2026-10-04T09:30:00Z").unwrap().to_rfc3339(),
+            "2026-10-04T09:30:00+00:00"
+        );
+        assert_eq!(
+            parse_since("2026-10-04 09:30:00").unwrap().to_rfc3339(),
+            "2026-10-04T09:30:00+00:00"
+        );
+        assert!(parse_since("nonsense").is_err());
+    }
+
+    #[test]
+    fn parse_message_ts_handles_export_formats() {
+        assert_eq!(
+            parse_message_ts("2026-01-01T00:00:00")
+                .unwrap()
+                .to_rfc3339(),
+            "2026-01-01T00:00:00+00:00"
+        );
+        assert!(parse_message_ts("2026-01-01T00:00:00.123000+00:00").is_some());
+        assert!(parse_message_ts("garbage").is_none());
+    }
 }