diff --git a/docs/config.md b/docs/config.md
index 1f8a54d2..57625622 100644
--- a/docs/config.md
+++ b/docs/config.md
@@ -32,7 +32,7 @@ You can also entirely disable the cache by setting `cache-duration` to `0`. When
passed on to the connected subscribers, but never stored on disk or even kept in memory longer than is needed to forward
the message to the subscribers.
-Subscribers can retrieve cached messaging using the [`poll=1` parameter](subscribe/api.md#polling), as well as the
+Subscribers can retrieve cached messaging using the [`poll=1` parameter](subscribe/api.md#polling-for-messages), as well as the
[`since=` parameter](subscribe/api.md#fetching-cached-messages).
## Behind a proxy (TLS, etc.)
diff --git a/docs/publish.md b/docs/publish.md
index 5ffaaf78..0e4a04d1 100644
--- a/docs/publish.md
+++ b/docs/publish.md
@@ -332,6 +332,85 @@ them with a comma, e.g. `tag1,tag2,tag3`.
Detail view of notifications with tags
+## Scheduled delivery
+You can delay the delivery of messages and let ntfy send them at a later date. This can be used to send yourself
+reminders or even to execute commands at a later date (if your subscriber acts on messages).
+
+Usage is pretty straight forward. You can set the delivery time using the `X-Delay` header (or any of its aliases: `Delay`,
+`X-At`, `At`, `X-In` or `In`), either by specifying a Unix timestamp (e.g. `1639194738`), a duration (e.g. `30m`,
+`3h`, `2 days`), or a natural language time string (e.g. `10am`, `8:30pm`, `tomorrow, 3pm`, `Tuesday, 7am`,
+[and more](https://github.com/olebedev/when)).
+
+As of today, the minimum delay you can set is **10 seconds** and the maximum delay is **3 days**. This can currently
+not be configured otherwise ([let me know](https://github.com/binwiederhier/ntfy/issues) if you'd like to change
+these limits).
+
+For the purposes of [message caching](config.md#message-cache), scheduled messages are kept in the cache until 12 hours
+after they were delivered (or whatever the server-side cache duration is set to). For instance, if a message is scheduled
+to be delivered in 3 days, it'll remain in the cache for 3 days and 12 hours. Also note that naturally,
+[turning off server-side caching](#message-caching) is not possible in combination with this feature.
+
+=== "Command line (curl)"
+ ```
+ curl -H "At: tomorrow, 10am" -d "Good morning" ntfy.sh/hello
+ curl -H "In: 30min" -d "It's 30 minutes later now" ntfy.sh/reminder
+ curl -H "Delay: 1639194738" -d "Unix timestamps are awesome" ntfy.sh/itsaunixsystem
+ ```
+
+=== "HTTP"
+ ``` http
+ POST /hello HTTP/1.1
+ Host: ntfy.sh
+ At: tomorrow, 10am
+
+ Good morning
+ ```
+
+=== "JavaScript"
+ ``` javascript
+ fetch('https://ntfy.sh/hello', {
+ method: 'POST',
+ body: 'Good morning',
+ headers: { 'At': 'tomorrow, 10am' }
+ })
+ ```
+
+=== "Go"
+ ``` go
+ req, _ := http.NewRequest("POST", "https://ntfy.sh/hello", strings.NewReader("Good morning"))
+ req.Header.Set("At", "tomorrow, 10am")
+ http.DefaultClient.Do(req)
+ ```
+
+=== "PHP"
+ ``` php-inline
+ file_get_contents('https://ntfy.sh/backups', false, stream_context_create([
+ 'http' => [
+ 'method' => 'POST',
+ 'header' =>
+ "Content-Type: text/plain\r\n" .
+ "At: tomorrow, 10am",
+ 'content' => 'Good morning'
+ ]
+ ]));
+ ```
+
+Here are a few examples (assuming today's date is **12/10/2021, 9am, Eastern Time Zone**):
+
+
+
+
+ Delay/At/In header | Message will be delivered at | Explanation |
+ 30m | 12/10/2021, 9:30am | 30 minutes from now |
+ 2 hours | 12/10/2021, 11:30am | 2 hours from now |
+ 1 day | 12/11/2021, 9am | 24 hours from now |
+ 10am | 12/10/2021, 10am | Today at 10am (same day, because it's only 9am) |
+ 8am | 12/11/2021, 8am | Tomorrow at 8am (because it's 9am already) |
+ 1639152000 | 12/10/2021, 11am (EST) | Today at 11am (EST) |
+
+ |
+
+
## Advanced features
### Message caching
@@ -347,7 +426,7 @@ client-side network disruptions, but arguably this feature also may raise privac
To avoid messages being cached server-side entirely, you can set `X-Cache` header (or its alias: `Cache`) to `no`.
This will make sure that your message is not cached on the server, even if server-side caching is enabled. Messages
are still delivered to connected subscribers, but [`since=`](subscribe/api.md#fetching-cached-messages) and
-[`poll=1`](subscribe/api.md#polling) won't return the message anymore.
+[`poll=1`](subscribe/api.md#polling-for-messages) won't return the message anymore.
=== "Command line (curl)"
```
@@ -393,7 +472,7 @@ are still delivered to connected subscribers, but [`since=`](subscribe/api.md#fe
]));
```
-### Firebase
+### Disable Firebase
!!! info
If `Firebase: no` is used and [instant delivery](subscribe/phone.md#instant-delivery) isn't enabled in the Android
app (Google Play variant only), **message delivery will be significantly delayed (up to 15 minutes)**. To overcome
diff --git a/docs/subscribe/api.md b/docs/subscribe/api.md
index b4449cf2..4a0bb858 100644
--- a/docs/subscribe/api.md
+++ b/docs/subscribe/api.md
@@ -239,7 +239,7 @@ or `all` (all cached messages).
curl -s "ntfy.sh/mytopic/json?since=10m"
```
-### Polling
+### Polling for messages
You can also just poll for messages if you don't like the long-standing connection using the `poll=1`
query parameter. The connection will end after all available messages have been read. This parameter can be
combined with `since=` (defaults to `since=all`).
@@ -248,6 +248,16 @@ combined with `since=` (defaults to `since=all`).
curl -s "ntfy.sh/mytopic/json?poll=1"
```
+### Fetching scheduled messages
+Messages that are [scheduled to be delivered](../publish.md#scheduled-delivery) at a later date are not typically
+returned when subscribing via the API, which makes sense, because after all, the messages have technically not been
+delivered yet. To also return scheduled messages from the API, you can use the `scheduled=1` (alias: `sched=1`)
+parameter (makes most sense with the `poll=1` parameter):
+
+```
+curl -s "ntfy.sh/mytopic/json?poll=1&sched=1"
+```
+
### Subscribing to multiple topics
It's possible to subscribe to multiple topics in one HTTP call by providing a
comma-separated list of topics in the URL. This allows you to reduce the number of connections you have to maintain:
diff --git a/server/server.go b/server/server.go
index b5bd31ee..6fadd9d9 100644
--- a/server/server.go
+++ b/server/server.go
@@ -344,20 +344,20 @@ func (s *Server) parseHeaders(header http.Header, m *message) (cache bool, fireb
m.Tags = append(m.Tags, strings.TrimSpace(s))
}
}
- whenStr := readHeader(header, "x-at", "at", "x-in", "in", "x-delay", "delay")
- if whenStr != "" {
+ delayStr := readHeader(header, "x-delay", "delay", "x-at", "at", "x-in", "in")
+ if delayStr != "" {
if !cache {
return false, false, errHTTPBadRequest
}
- at, err := util.ParseFutureTime(whenStr, time.Now())
+ delay, err := util.ParseFutureTime(delayStr, time.Now())
if err != nil {
return false, false, errHTTPBadRequest
- } else if at.Unix() < time.Now().Add(s.config.MinDelay).Unix() {
+ } else if delay.Unix() < time.Now().Add(s.config.MinDelay).Unix() {
return false, false, errHTTPBadRequest
- } else if at.Unix() > time.Now().Add(s.config.MaxDelay).Unix() {
+ } else if delay.Unix() > time.Now().Add(s.config.MaxDelay).Unix() {
return false, false, errHTTPBadRequest
}
- m.Time = at.Unix()
+ m.Time = delay.Unix()
}
return cache, firebase, nil
}
diff --git a/util/time.go b/util/time.go
index ed68b766..70501210 100644
--- a/util/time.go
+++ b/util/time.go
@@ -14,6 +14,8 @@ var (
durationStrRegex = regexp.MustCompile(`(?i)^(\d+)\s*(d|days?|h|hours?|m|mins?|minutes?|s|secs?|seconds?)$`)
)
+// ParseFutureTime parses a date/time string to a time.Time. It supports unix timestamps, durations
+// and natural language dates
func ParseFutureTime(s string, now time.Time) (time.Time, error) {
s = strings.TrimSpace(s)
t, err := parseUnixTime(s, now)