ClickBench VS prql

Benchmarks: https://benchmark.clickhouse.com

So definitely compared against PostgreSQL, MariaDB it is significantly faster.

On par with lower-end Snowflake.

TimescaleDB primarily serves operational use cases: Developers building products on top of live data, where you are regularly streaming in fresh data, and you often know what many queries look like a priori, because those are powering your live APIs, dashboards, and product experience.

That's different from a data warehouse or many traditional "OLAP" use cases, where you might dump a big dataset statically, and then people will occasionally do ad-hoc queries against it. This is the big weather dataset file sitting on your desktop that you occasionally query while on holidays.

So it's less about "can you store weather data", but what does that use case look like? How are the queries shaped? Are you saving a single dataset for ad-hoc queries across the entire dataset, or continuously streaming in new data, and aging out or de-prioritizing old data?

In most of the products we serve, customers are often interested in recent data in a very granular format ("shallow and wide"), or longer historical queries along a well defined axis ("deep and narrow").

For example, this is where the benefits of TimescaleDB's segmented columnar compression emerges. It optimizes for those queries which are very common in your application, e.g., an IoT application that groups by or selected by deviceID, crypto/fintech analysis based on the ticker symbol, product analytics based on tenantID, etc.

If you look at Clickbench, what most of the queries say are: Scan ALL the data in your database, and GROUP BY one of the 100 columns in the web analytics logs.

- https://github.com/ClickHouse/ClickBench/blob/main/clickhous...

There are almost no time-predicates in the benchmark that Clickhouse created, but perhaps that is not surprising given it was designed for ad-hoc weblog analytics at Yandex.

So yes, Timescale serves many products today that use weather data, but has made different choices than Clickhouse (or things like DuckDB, pg_analytics, etc) to serve those more operational use cases.

We tested with 43 Clickbench SQL queries. Queries on the Variant columns are about 10% slower than those on pre-defined static columns, and 8 times faster than those on JSON columns. (For I/O reasons, most cold runs on JSONB data failed with OOM.)

I have a project dedicated to this topic: https://github.com/ClickHouse/ClickBench

It is important to explain the limitations of a benchmark, provide a methodology, and make it reproducible. It also has to be simple enough, otherwise it will not be realistic to include a large number of participants.

I'm also collecting all database benchmarks I could find: https://github.com/ClickHouse/ClickHouse/issues/22398

Indeed, ClickHouse results were run on an older instance type of the same family and size (c5.4xlarge for ClickHouse and c6a.4xlarge for Timescale), so if anything ClickHouse results are at a slight disadvantage.

This is an open source benchmark - we'd love contributions from Timescale enthusiasts if we missed something: https://github.com/ClickHouse/ClickBench/

Interesting project! Thank you for open sourcing and sharing. Agree that local and embedded analytics are an increasing trend, I see it too.

A couple of questions:

* I’m curious what the difficulties were in the implementation. I suspect it is quite a challenge to implement this support in the current SQLite architecture, and would curious to know which parts were tricky and any design trade-off you were faced with.

* Aside from ease-of-use (install extension, no need for a separate analytical database system), I wonder if there are additional benefits users can anticipate resulting from a single system architecture vs running an embedded OLAP store like DuckDB or clickhouse-local / chdb side-by-side with SQLite? Do you anticipate performance or resource efficiency gains, for instance?

* I am also curious, what the main difficulty with bringing in a separate analytical database is, assuming it natively integrates with SQLite. I may be biased, but I doubt anything can approach the performance of native column-oriented systems, so I'm curious what the tipping point might be for using this extension vs using an embedded OLAP store in practice.

Btw, would love for you or someone in the community to benchmark Stanchion in ClickBench and submit results! (https://github.com/ClickHouse/ClickBench/)

Disclaimer: I work on ClickHouse.

> I completely attribute this to SQL being difficult or "backwards" to parse. I mean backwards in the way that in SQL you start with what you want first (the SELECT) rather than what you have and widdling it down.

> The turning point for me was to just accept SQL for what it is.

Or just write PRQL and compile it to SQL

https://github.com/PRQL/prql

Hey HN! We’ve built Pretzel, an open-source data exploration and visualization tool that runs fully in the browser and can handle large files (200 MB CSV on my 8gb MacBook air is snappy). It’s also reactive - so if, for example, you change a filter, all the data transform blocks after it re-evaluate automatically. You can try it here: https://pretzelai.github.io/ (static hosted webpage) or see a demo video here: https://www.youtube.com/watch?v=73wNEun_L7w

You can play with the demo CSV that’s pre-loaded (GitHub data of text-editor adjacent projects) or upload your own CSV/XLSX file. The tool runs fully in-browser—you can disconnect from the internet once the website loads—so feel free to use sensitive data if you like.

Here’s how it works: You upload a CSV file and then, explore your data as a series of successive data transforms and plots. For example, you might: (1) Remove some columns; (2) Apply some filters (remove nulls, remove outliers, restrict time range etc); (3) Do a pivot (i.e, a group-by but fancier); (4) Plot a chart; (5) Download the chart and the the transformed data. See screenshot: https://imgur.com/a/qO4yURI

In the UI, each transform step appears as a “Block”. You can always see the result of the full transform in a table on the right. The transform blocks are editable - for instance in the example above, you can go to step 2, change some filters and the reactivity will take care of re-computing all the cells that follow, including the charts.

We wanted Pretzel to run locally in the browser and be extremely performant on large files. So, we parse CSVs with the fastest CSV parser (uDSV: https://github.com/leeoniya/uDSV) and use DuckDB-Wasm (https://github.com/duckdb/duckdb-wasm) to do all the heavy lifting of processing the data. We also wanted to allow for chained data transformations where each new block operates on the result of the previous block. For this, we’re using PRQL (https://prql-lang.org/) since it maps 1-1 with chained data transform blocks - each block maps to a chunk of PRQL which when combined, describes the full data transform chain. (PRQL doesn’t support DuckDB’s Pivot statement though so we had to make some CTE based hacks).

There’s also an AI block: This is the only (optional) feature that requires an internet connection but we’re working on adding local model support via Ollama. For now, you can use your own OpenAI API key or use an AI server we provide (GPT4 proxy; it’s loaded with a few credits), specify a transform in plain english and get back the SQL for the transform which you can edit.

Our roadmap includes allowing API calls to create new columns; support for an SQL block with nice autocomplete features, and a Python block (using Pyodide to run Python in the browser) on the results of the data transforms, much like a jupyter notebook.

There’s two of us and we’ve only spent about a week coding this and fixing major bugs so there are still some bugs to iron out. We’d love for you to try this and to get your feedback!

> Looks like PRQL doesn't have a Go library so I guess they just really wanted something in Go?

There's some C bindings and the example in the README shows integration with Go:

https://github.com/PRQL/prql/tree/main/prqlc/bindings/prqlc-...

Can someone tell me why PRQL is better? I went here: https://github.com/PRQL/prql

It looks nice, but what's the strengths compared to SQL?

PRQL [1] is a compile-to-SQL relational querying language that puts FROM first.

[1] https://prql-lang.org

https://prql-lang.org/ might be an answer for this. As a cross-database pipelined language, it would allow RAG to be intermixed with the query, and the syntax may(?) be more reliable to generate