Query your data

With Timescale, there's no need to learn a custom query language, because Timescale supports full SQL. You can use your SQL knowledge along with the rich ecosystem of PostgreSQL tools, and add the extra features and performance of Timescale.

Here are some query examples so you can get familiar with using SQL alongside popular Timescale functions.

Many of the queries below show a filter for the last four days of data. This accounts for the nuance of stock trade data which only occurs Monday to Friday on the New York Stock Exchange.

If you load the provided data on a Monday, the most recent data is from Friday afternoon. Therefore, selecting data for the last day or two would return no results.

You can adjust the time frame based on the data that you downloaded and to explore other time-ranges in the provided data.

To select all the stock data from the previous four days, use the WHERE clause to filter the result using a relative time interval. This example uses an interval of four days, so data is displayed even if you run this on a weekend or a Monday:

SELECT * FROM stocks_real_time srt
WHERE time > now() - INTERVAL '4 days';

Copy

Use the ORDER BY clause to define the order of results from your query. With stock trade data, there are often multiple trades each second for popular stocks like Amazon. Therefore, you cannot order data descending by the time alone. This is a common problem with high-frequency data like stocks, crypto, and IoT metrics. You need to order the results by additional information to correctly display the order in which trades were made with the exchange.

For the stocks_real_time data, the day_volume column serves as additional information to help you order the trades correctly, even when there are multiple trades per second. The day_volume value increases by the number of stocks traded with each tick.

SELECT * FROM stocks_real_time srt
WHERE symbol='AMZN'
ORDER BY time DESC, day_volume desc
LIMIT 10;

Copy

The results look like this:

time                         |symbol|price    |day_volume|
-----------------------------+------+---------+----------+
2022-05-04 14:11:32.000 -0400|AMZN  |2429.1191|   3134115|
2022-05-04 14:11:28.000 -0400|AMZN  |  2428.53|   3133809|
2022-05-04 14:11:28.000 -0400|AMZN  |  2428.53|   3133644|
2022-05-04 14:11:28.000 -0400|AMZN  |  2428.53|   3133638|
2022-05-04 14:11:28.000 -0400|AMZN  |  2428.53|   3133602|
2022-05-04 14:11:18.000 -0400|AMZN  |  2426.83|   3132536|
2022-05-04 14:11:18.000 -0400|AMZN  |  2426.83|   3132009|
2022-05-04 14:11:18.000 -0400|AMZN  |  2426.83|   3131887|
2022-05-04 14:11:18.000 -0400|AMZN  |  2426.83|   3131848|
2022-05-04 14:11:18.000 -0400|AMZN  |  2426.83|   3131844|

Copy

There are multiple trades every second, but you know that the order of trades is correct because the day_volume column is ordered correctly.

Use the avg() function with a WHERE clause to only include trades for Apple stock within the last 4 days. You can use the JOIN operator to fetch results based on the name of a company instead of the symbol.

SELECT
    avg(price)
FROM stocks_real_time srt
JOIN company c ON c.symbol = srt.symbol
WHERE c.name = 'Apple' AND time > now() - INTERVAL '4 days';

Copy

Timescale has many custom-built SQL functions to help you perform time-series analysis in fewer lines of code. Here's how to use three of these functions:

• first(): find the earliest value based on a time within an aggregate group
• last(): find the latest value based on time within an aggregate group
• time_bucket(): bucket data by arbitrary time intervals and calculate aggregates over those intervals

The first() and last() functions retrieve the first and last value of one column when ordered by another.

For example, the stock data has a timestamp column time and numeric column price. You can use first(price, time) to get the first value in the price column when ordered with respect to an increasing time column.

In this query, you use both the first() and last() functions to find the first and last trading price for each company for the last three days.

SELECT symbol, first(price,time), last(price, time)
FROM stocks_real_time srt
WHERE time > now() - INTERVAL '3 days'
GROUP BY symbol
ORDER BY symbol;

Copy

The results look like this:

symbol|first   |last    |
------+--------+--------+
AAPL  |  156.26|  160.79|
ABBV  |  145.38|  150.32|
ABNB  |  152.08|  148.05|
ABT   |   113.5|  112.88|
ADBE  |   391.2|  403.94|
AMAT  |  109.72|113.0464|
AMD   |  84.938|  93.585|
AMGN  |   233.3|  233.11|
...   |   ...  |    ... |

Copy

The time_bucket() function enables you to take a time column and "bucket" the values based on an interval of your choice. Typically, you bucket time so that you can perform an aggregation over the chosen interval.

For example, consider a table that records incrementing values every hour. To aggregate the daily totals of the values, you can use the time_bucket() function on the time column to bucket the hourly data into daily data and then perform a sum() on the value column to get the total sum of your values across each day.

Expand image

For more information on the time_bucket() function, see the API documentation.

To see time_bucket() in action with the stock trade data, you can calculate the average daily price of each trading symbol over the last week.

Use the time_bucket() function with an interval of 1-day, the avg() function on the price data, select the symbol, and finally GROUP BY the bucket and symbol. The WHERE limits the results to days within the last week. Finally, the ORDER BY clause orders the results first on the bucketed date, then by symbol.

SELECT
  time_bucket('1 day', time) AS bucket,
  symbol,
  avg(price)
FROM stocks_real_time srt
WHERE time > now() - INTERVAL '1 week'
GROUP BY bucket, symbol
ORDER BY bucket, symbol;

Copy

The results look like this:

bucket                       |symbol|avg               |
-----------------------------+------+------------------+
2022-04-26 20:00:00.000 -0400|AAPL  |157.16595920217668|
2022-04-26 20:00:00.000 -0400|ABBV  | 157.8470588235293|
2022-04-26 20:00:00.000 -0400|ABNB  |152.33858034970868|
2022-04-26 20:00:00.000 -0400|ABT   |117.13218965517241|
2022-04-26 20:00:00.000 -0400|ADBE  |398.63256560534745|
2022-04-26 20:00:00.000 -0400|AMAT  |108.92946602133563|

Copy

In these results, you might notice that the bucket column, which represents a time_bucket() of one week, starts on the beginning date of the bucket, not the current time that you run the query. To learn more about how time buckets are calculated, see the how-to guide for time buckets.

Now that you're familiar with some Timescale queries and functions, like time_bucket, learn about continuous aggregates in the next section.

For more information about the functions provided by Timescale and Timescale Toolkit extension, see the hyperfunctions section.