For instance, if the user query is:
SELECT COUNT(*) FROM my_vtab;
... the query your virtual table will effectively see is:
SELECT * FROM my_vtab;
SQLite does the counting. That's great, unless you already know the count and could have reported it directly rather than actually returning every row in the table. You're forced to retrieve and return every row because you have no idea that it was actually just a count.
As another example, if the user query includes a join, you won't see the join. Instead, you will receive a series of N queries for individual IDs, even if you could have more efficiently retrieved them in a batch.
The join one is particularly nasty. If you're writing a virtual table that accesses a remote resource with some latency, any join will absolutely ruin your performance as you pay a full network roundtrip for each of those N queries.
I wrote a module that exposes remote SQL Server/PostgreSQL/MySQL servers as SQLite virtual tables, and joins basically don't work at all if your server is not on your local network. There's nothing I can do about it (other than heuristically guessing what IDs might be coming and request them ahead of time) because SQLite doesn't provide enough information to the virtual table layer. It's my understanding that PostgreSQL's foreign data wrappers (a similar feature to SQLite's virtual tables) push much more information about the query down to the wrapper layer, but I haven't used it myself.
Is it possible to wait until all of the queries are in and then do some query planning at your end to resolve this? Or is it entirely synchronous with no way to do that?
It really should be, as the code is tiny and this functionality is not overly exotic.
The CSV virtual table source code is a good pedagogical tool for teaching how to build a virtual table, though.
[1]: https://cldellow.com/2018/06/22/sqlite-parquet-vtable.html
Is csvw with linked data URIs also doable?
Zsvlib on cloudfuzz would be good if that's not already
Yeah linked data schema support is distinct from the parser primitives and xsd data type uris, for example.
- "Comparing SQLite, DuckDB and Arrow with UN trade data" (2021) https://news.ycombinator.com/item?id=29010103 ; partial benchmarks of query time and RAM requirements [relative to data size] would be
- "Introducing Apache Arrow Flight SQL: Accelerating Database Access" (2022) https://arrow.apache.org/blog/2022/02/16/introducing-arrow-f... :
> Motivation: While standards like JDBC and ODBC have served users well for decades, they fall short for databases and clients which wish to use Apache Arrow or columnar data in general. Row-based APIs like JDBC or PEP 249 require transposing data in this case, and for a database which is itself columnar, this means that data has to be transposed twice—once to present it in rows for the API, and once to get it back into columns for the consumer. Meanwhile, while APIs like ODBC do provide bulk access to result buffers, this data must still be copied into Arrow arrays for use with the broader Arrow ecosystem, as implemented by projects like Turbodbc. Flight SQL aims to get rid of these intermediate steps.
## "The Virtual Table Mechanism Of SQLite" https://sqlite.org/vtab.html :
> - One cannot create a trigger on a virtual table.
Just posted about eBPF a few days ago; opcodes have costs that are or are not costed: https://news.ycombinator.com/item?id=31688180
> - One cannot create additional indices on a virtual table. (Virtual tables can have indices but that must be built into the virtual table implementation. Indices cannot be added separately using CREATE INDEX statements.)
It looks like e.g. sqlite-parquet-vtable implements shadow tables to memoize row group filters. How does JOIN performance vary amongst sqlite virtual table implementations?
> - One cannot run ALTER TABLE ... ADD COLUMN commands against a virtual table.
Are there URIs in the schema? Mustn't there thus be a meta-schema that does e.g. nested structs with portable types [with URIs], (and jsonschema, [and W3C SHACL])? #nbmeta #linkedresearch
## /? sqlite arrow virtual table
- sqlite-parquet-vtable reads parquet with arrow for SQLite virtual tables https://github.com/cldellow/sqlite-parquet-vtable :
$ sqlite/sqlite3
sqlite> .eqp on
sqlite> .load build/linux/libparquet
sqlite> CREATE VIRTUAL TABLE demo USING parquet('parquet-generator/99-rows-1.parquet');
sqlite> SELECT * FROM demo;
//
sqlite> SELECT * FROM demo WHERE foo = 123;
sqlite> SELECT * FROM demo WHERE foo = '123'; // incurs a severe query plan performance regression without immediate feedback
`EXPLAIN QUERY PLAN` https://www.sqlite.org/eqp.html :
> The EXPLAIN QUERY PLAN SQL command is used to obtain a high-level description of the strategy or plan that SQLite uses to implement a specific SQL query. Most significantly, EXPLAIN QUERY PLAN reports on the way in which the query uses database indices. This document is a guide to understanding and interpreting the EXPLAIN QUERY PLAN output. [...] Table and Index Scans [...] Temporary Sorting B-Trees (when there's not an `INDEX` for those columns) ... `.eqp on`
The SQLite "Query Planner" docs https://www.sqlite.org/queryplanner.html list Big-O computational complexity bound estimates for queries with and without prexisting indices.
## database / csv benchmarks