{ "id": "ts.nrevrange", "title": "TS.NREVRANGE", "url": "https://redis.io/docs/latest/commands/ts.nrevrange/", "summary": "Query a range across multiple time series in reverse direction, returning the results grouped by timestamp", "tags": [ "docs", "develop", "stack", "oss", "rs", "rc", "oss", "kubernetes", "clients" ], "last_updated": "2026-07-30T08:54:30-07:00", "page_type": "content", "content_hash": "83b028dc6d024d939e370ca448a110998dab84dbbe0824bf0a1757efff67fee6", "sections": [ { "id": "overview", "title": "Overview", "role": "overview", "text": "Query a range across an explicit list of time series in the reverse direction. `TS.NREVRANGE` groups results by timestamp: for each timestamp, it returns the timestamp followed by one value per time series, in the order the keys are given. With `AGGREGATION`, a time series can contribute more than one value per timestamp — one for each aggregator requested. A time series with no sample at a given timestamp returns `NaN` for that timestamp.\n\n\nIn a Redis cluster, all specified keys must map to the same hash slot. `TS.NREVRANGE` is a [single hash slot](https://redis.io/docs/latest/operate/oss_and_stack/reference/cluster-spec#key-distribution-model) command; it does not split a request across shards or merge replies from multiple hash slots.\n\n\n[Examples](#examples)" }, { "id": "required-arguments", "title": "Required arguments", "role": "content", "text": "
\nnumkeys\n\nis the number of time series keys that follow. It must be a positive integer and must equal the number of `key` arguments.\n
\n\n
\nkey [key ...]\n\nare the explicit time series keys to query. Key order and duplicate keys are significant: the output has one value per `key` argument, in the order the keys are given. A repeated key produces a separate value for each occurrence.\n
\n\n
\nfromTimestamp\n\nis the start timestamp for the range query (integer Unix timestamp in milliseconds) or `-` to denote the timestamp of the earliest sample among the specified time series. The range is inclusive.\n
\n\n
\ntoTimestamp\n\nis the end timestamp for the range query (integer Unix timestamp in milliseconds) or `+` to denote the timestamp of the latest sample among the specified time series. The range is inclusive.\n
" }, { "id": "optional-arguments", "title": "Optional arguments", "role": "parameters", "text": "
\nLATEST\n\nis used when a time series is a compaction. With `LATEST`, TS.NREVRANGE also reports the compacted value of the latest, possibly partial, bucket, given that this bucket's start time falls within `[fromTimestamp, toTimestamp]`. Without `LATEST`, TS.NREVRANGE does not report the latest, possibly partial, bucket. When a time series is not a compaction, `LATEST` is ignored.\n\nThe data in the latest bucket of a compaction is possibly partial. A bucket is _closed_ and compacted only upon arrival of a new sample that _opens_ a new _latest_ bucket. There are cases, however, when the compacted value of the latest, possibly partial, bucket is also required. In such a case, use `LATEST`.\n
\n\n
\nFILTER_BY_TS ts...\n\nfilters samples by a list of specific timestamps. A sample passes the filter if its exact timestamp is specified and falls within `[fromTimestamp, toTimestamp]`. Samples are filtered before pivoting.\n\nWhen used together with `AGGREGATION`: samples are filtered before being aggregated.\n
\n\n
\nFILTER_BY_VALUE min max\n\nfilters samples by minimum and maximum values. A sample passes the filter if its value is within the inclusive range `[min, max]`. `min` and `max` cannot be NaN values. Samples are filtered before pivoting.\n\nWhen used together with `AGGREGATION`: samples are filtered before being aggregated.\n
\n\n
\nCOUNT count\n\nlimits the number of returned entries, keeping those with the highest timestamps. The limit is applied after the per-timestamp merge.\n
\n\n
\nALIGN align\n\nis a time bucket alignment control for `AGGREGATION`. It controls the time bucket timestamps by changing the reference timestamp on which a bucket is defined.\n\n`align` values include:\n\n - `start` or `-`: The reference timestamp will be the query start interval time (`fromTimestamp`) which can't be `-`\n - `end` or `+`: The reference timestamp will be the query end interval time (`toTimestamp`) which can't be `+`\n - A specific timestamp: align the reference timestamp to a specific time\n\nNote: When not provided, alignment is set to `0`.\n
\n\n
\nAGGREGATION aggregators [aggregators ...] bucketDuration\n\naggregates samples into time buckets.\n\nProvide one or more aggregators per key, as a separate argument in the same order as the keys: the first argument applies to the first key, and so on. The number of these arguments must equal `numkeys`, and all keys share the same `bucketDuration`.\n\nEach per-key argument is either a single aggregator or a comma-separated list of aggregators (for example `avg,max`), exactly as in [`TS.REVRANGE`](https://redis.io/docs/latest/commands/ts.revrange/); no whitespace is allowed. A key contributes one value for each aggregator you list for it, and a key's values appear together in the reply, in the order you list its aggregators. To compute several aggregations for one series, give that key a comma-separated list such as `min,max`.\n\n - each `aggregator` is one of the following:\n\n | aggregator | Description |\n | ------------ | --------------------------------------------------------------- |\n | `avg` | Arithmetic mean of all non-NaN values |\n | `sum` | Sum of all non-NaN values |\n | `min` | Minimum non-NaN value |\n | `max` | Maximum non-NaN value |\n | `range` | Difference between the maximum and the minimum non-NaN values |\n | `count` | Number of non-NaN values |\n | `countNaN` | Number of NaN values |\n | `countAll` | Number of values, including NaN and non-NaN |\n | `first` | The non-NaN value with the lowest timestamp in the bucket |\n | `last` | The non-NaN value with the highest timestamp in the bucket |\n | `std.p` | Population standard deviation of the non-NaN values |\n | `std.s` | Sample standard deviation of the non-NaN values |\n | `var.p` | Population variance of the non-NaN values |\n | `var.s` | Sample variance of the non-NaN values |\n | `twa` | Time-weighted average over the bucket's timeframe (ignores NaN values) |\n\n - `bucketDuration` is the duration of each bucket, in milliseconds.\n\n Without `ALIGN`, bucket start times are multiples of `bucketDuration`.\n\n With `ALIGN align`, bucket start times are multiples of `bucketDuration` with remainder `align % bucketDuration`.\n\n The first bucket start time is less than or equal to `fromTimestamp`.\n
\n\n
\n[BUCKETTIMESTAMP bt]\n\ncontrols how bucket timestamps are reported.\n\n| `bt` | Timestamp reported for each bucket |\n| ---------------- | ------------------------------------------------------------- |\n| `-` or `start` | the bucket's start time (default) |\n| `+` or `end` | the bucket's end time |\n| `~` or `mid` | the bucket's mid time (rounded down if not an integer) |\n
\n\n
\n[EMPTY]\n\nis a flag, which, when specified, reports aggregations also for empty buckets.\n\n| aggregator | Value reported for each empty bucket |\n| -------------------- | ------------------------------------ |\n| `sum`, `count` | `0` |\n| `last` | The value of the last sample before the bucket's start. `NaN` when no such sample. |\n| `twa` | Average value over the bucket's timeframe based on linear interpolation of the last sample before the bucket's start and the first sample after the bucket's end. `NaN` when no such samples. |\n| `min`, `max`, `range`, `avg`, `first`, `std.p`, `std.s` | `NaN` |\n\nRegardless of the values of `fromTimestamp` and `toTimestamp`, no data is reported for buckets that end before the earliest sample or begin after the latest sample in the time series. When a bucket is reported for one key but a different key has no data in that bucket, that key's value is `NaN`.\n
" }, { "id": "examples", "title": "Examples", "role": "example", "text": "
\nPivot raw samples from multiple series by timestamp\n\nCreate two time series and add samples at partially overlapping timestamps.\n\n[code example]\n\nQuery both series and group the samples by timestamp. One entry is returned for every distinct timestamp produced by at least one key, ordered from the highest timestamp to the lowest, with the values ordered by input key. A key with no sample at that timestamp has a `NaN` value.\n\n[code example]\n
\n\n
\nAggregate each series with a per-key aggregator\n\nIn aggregation mode, supply one aggregator per key, in key order. Here `{sensor}:1` is aggregated with `avg` and `{sensor}:2` with `sum`, both over 1000-millisecond buckets.\n\n[code example]\n
\n\n
\nApply multiple aggregators to a series\n\nTo compute several aggregators for a single key, pass them as a comma-separated list. Here `{sensor}:1` uses `avg,max` (two values) and `{sensor}:2` uses `sum` (one value). Each timestamp's values are a single flat list: `{sensor}:1`'s `avg`, then its `max`, then `{sensor}:2`'s `sum`.\n\n[code example]\n
" }, { "id": "details", "title": "Details", "role": "content", "text": "" }, { "id": "semantics", "title": "Semantics", "role": "content", "text": "`TS.NREVRANGE` behaves like running a compatible [`TS.REVRANGE`](https://redis.io/docs/latest/commands/ts.revrange/) over each input key and then performing a server-side outer join by timestamp. Rows are returned from the highest timestamp to the lowest.\n\nIn raw mode (no `AGGREGATION`):\n\n- One entry is returned for every distinct timestamp produced by at least one key.\n- If a key has no sample at that timestamp, that key's value is `NaN`.\n\nIn aggregation mode (with `AGGREGATION`):\n\n- One aggregation applies to each key, in key order; the number of these arguments must equal `numkeys`, and all keys share one `bucketDuration`.\n- Each per-key argument is a single aggregator or a comma-separated list of aggregators, and a key contributes one value per aggregator.\n- Each timestamp's values form a single flat list, ordered by key and, within each key, by the order its aggregators are listed. For example, `avg,max` for the first key and `sum` for the second produce values of the form `[avg, max, sum]`.\n- When a key has no data for a given bucket, all of that key's values are `NaN`.\n- With `EMPTY`, empty buckets can produce rows in which every value is `NaN`." }, { "id": "nan-values", "title": "NaN values", "role": "content", "text": "A `NaN` value can mean that a key had no sample at that timestamp or no samples at that time bucket, or that the key stored or aggregated to a real `NaN`. These two cases are indistinguishable in the reply.\n\n| Case | Value |\n| ----------------------------------------------- | ----------------------------------------------- |\n| Key has a sample at that timestamp | The sample value |\n| Key has no sample at that timestamp | `NaN` |\n| Key has aggregated data for that time bucket | The aggregated value |\n| Key has no samples at that time bucket | `NaN` |\n| Key stores or aggregates to a real `NaN` | `NaN`, indistinguishable from no data |" }, { "id": "redis-software-and-redis-cloud-compatibility", "title": "Redis Software and Redis Cloud compatibility", "role": "content", "text": "| Redis
Software | Redis
Cloud | Notes |\n|:----------------------|:-----------------|:------|\n| ❌ Not supported
| ❌ Flexible & Annual
❌ Free & Fixed |" }, { "id": "return-information", "title": "Return information", "role": "returns", "text": "**RESP2:**\n\nOne of the following:\n* [Array reply](https://redis.io/docs/latest/develop/reference/protocol-spec#arrays) with one entry per timestamp, ordered by decreasing timestamp. Each entry is an [Array reply](https://redis.io/docs/latest/develop/reference/protocol-spec#arrays) composed of an [Integer reply](https://redis.io/docs/latest/develop/reference/protocol-spec#integers) (the timestamp) and a flat [Array reply](https://redis.io/docs/latest/develop/reference/protocol-spec#arrays) of [Simple string reply](https://redis.io/docs/latest/develop/reference/protocol-spec#simple-strings) values. The values are concatenated across keys in input order; with `AGGREGATION`, each key contributes one value per aggregator, otherwise one value per key. A missing value is reported as `NaN`. The reply is an empty array when no samples match.\n* [Simple error reply](https://redis.io/docs/latest/develop/reference/protocol-spec#simple-errors) in these cases: invalid arguments, wrong number of aggregators, unknown aggregation type, wrong key type, etc.\n\n**RESP3:**\n\nOne of the following:\n* [Array reply](https://redis.io/docs/latest/develop/reference/protocol-spec#arrays) with one entry per timestamp, ordered by decreasing timestamp. Each entry is an [Array reply](https://redis.io/docs/latest/develop/reference/protocol-spec#arrays) composed of an [Integer reply](https://redis.io/docs/latest/develop/reference/protocol-spec#integers) (the timestamp) and a flat [Array reply](https://redis.io/docs/latest/develop/reference/protocol-spec#arrays) of [Double reply](https://redis.io/docs/latest/develop/reference/protocol-spec#doubles) values. The values are concatenated across keys in input order; with `AGGREGATION`, each key contributes one value per aggregator, otherwise one value per key. A missing value is reported as `NaN`. The reply is an empty array when no samples match.\n* [Simple error reply](https://redis.io/docs/latest/develop/reference/protocol-spec#simple-errors) in these cases: invalid arguments, wrong number of aggregators, unknown aggregation type, wrong key type, etc." }, { "id": "see-also", "title": "See also", "role": "related", "text": "[`TS.NRANGE`](https://redis.io/docs/latest/commands/ts.nrange/) | [`TS.REVRANGE`](https://redis.io/docs/latest/commands/ts.revrange/) | [`TS.MREVRANGE`](https://redis.io/docs/latest/commands/ts.mrevrange/)" }, { "id": "related-topics", "title": "Related topics", "role": "related", "text": "[RedisTimeSeries](https://redis.io/docs/latest/develop/data-types/timeseries/)" } ], "examples": [ { "id": "examples-ex0", "language": "plaintext", "code": "127.0.0.1:6379> TS.CREATE {sensor}:1\nOK\n127.0.0.1:6379> TS.CREATE {sensor}:2\nOK\n127.0.0.1:6379> TS.MADD {sensor}:1 1000 10 {sensor}:1 2000 12\n1) (integer) 1000\n2) (integer) 2000\n127.0.0.1:6379> TS.MADD {sensor}:2 1000 20 {sensor}:2 3000 25\n1) (integer) 1000\n2) (integer) 3000", "section_id": "examples" }, { "id": "examples-ex1", "language": "plaintext", "code": "127.0.0.1:6379> TS.NREVRANGE 2 {sensor}:1 {sensor}:2 - +\n1) 1) (integer) 3000\n 2) 1) NaN\n 2) 25\n2) 1) (integer) 2000\n 2) 1) 12\n 2) NaN\n3) 1) (integer) 1000\n 2) 1) 10\n 2) 20", "section_id": "examples" }, { "id": "examples-ex2", "language": "plaintext", "code": "127.0.0.1:6379> TS.MADD {sensor}:1 1000 10 {sensor}:1 1100 20 {sensor}:1 2000 30\n1) (integer) 1000\n2) (integer) 1100\n3) (integer) 2000\n127.0.0.1:6379> TS.MADD {sensor}:2 1000 5 {sensor}:2 1100 15 {sensor}:2 2000 25\n1) (integer) 1000\n2) (integer) 1100\n3) (integer) 2000\n127.0.0.1:6379> TS.NREVRANGE 2 {sensor}:1 {sensor}:2 - + AGGREGATION avg sum 1000\n1) 1) (integer) 2000\n 2) 1) 30\n 2) 25\n2) 1) (integer) 1000\n 2) 1) 15\n 2) 20", "section_id": "examples" }, { "id": "examples-ex3", "language": "plaintext", "code": "127.0.0.1:6379> TS.NREVRANGE 2 {sensor}:1 {sensor}:2 - + AGGREGATION avg,max sum 1000\n1) 1) (integer) 2000\n 2) 1) 30\n 2) 30\n 3) 25\n2) 1) (integer) 1000\n 2) 1) 15\n 2) 20\n 3) 20", "section_id": "examples" } ] }