Replace sqlite3 with sqlean

[kracekumar]

Description

1. Sqlean is drop in replacement for sqlite3 with extra extensions.
2. It's built 12 extra extensions. From the website

Think of them as the extended standard library for SQLite:

    [crypto](https://antonz.org/sqlean-crypto/): Hashing, encoding and decoding data.
    [define](https://antonz.org/sqlean-define/): User-defined functions and dynamic SQL.
    [fileio](https://antonz.org/sqlean-fileio/): Reading and writing files and catalogs.
    [fuzzy](https://github.com/nalgeon/sqlean/blob/main/docs/fuzzy.md): Fuzzy string matching and phonetics.
    [ipaddr](https://github.com/nalgeon/sqlean/blob/main/docs/ipaddr.md): IP address manipulation.
    [math](https://github.com/nalgeon/sqlean/blob/main/docs/math.md): Math functions.
    [regexp](https://antonz.org/sqlean-regexp/): Pattern matching using regular expressions.
    [stats](https://github.com/nalgeon/sqlean/blob/main/docs/stats.md): Math statistics — median, percentiles, etc.
    [text](https://antonz.org/sqlean-text/): Powerful Unicode-aware string functions.
    [time](https://antonz.org/sqlean-time/): High-precision date/time.
    [uuid](https://github.com/nalgeon/sqlean/blob/main/docs/uuid.md): Universally Unique IDentifiers.
    [vsv](https://github.com/nalgeon/sqlean/blob/main/docs/vsv.md): CSV files as virtual tables.

3. This provides some extra extensions inside sqlite which can be handy as reported in writefile function with fileio #119
4. It's worth experimenting to see how it goes especially in terms of regression.
5. Note: I ran ruff formatter and it fixed a bunch of style issues.

Notes:

• I haven't used sqlean in production or in any other project not sure of any issues
• This may sound risky to replace sqlite3, though it's a drop-in replacement.

Checklist

• I've added this contribution to the CHANGELOG.md file.

[kracekumar]

All the formatting done by ruff.

[kracekumar]

Adding extra table and columns from sqlean define

[kracekumar]

Here is a simple benchmark

# -*- coding: utf-8 -*-

from litecli.main import LiteCli
import timeit


def perf():
    dbname=":memory:"
    cli = LiteCli()
    cli.connect(dbname)
    # Create a table
    cli.run_query("create table test (a text)")
    # Insert 10000 rows
    for i in range(10000):
        cli.run_query(f'insert into test values("{i}")')
    # Delete the table
    cli.run_query("drop table test")


def main():
    # timeit is not needed when using hyperfine
    print(timeit.timeit("perf()", globals=globals(), number=10))


if __name__ == "__main__":
    main()

Default Sqlite3

hyperfine --shell /opt/homebrew/bin/fish  --runs 5 --show-output  'uv run python litecli/performance.py'
Benchmark 1: uv run python litecli/performance.py
3.7350782089924905
3.7475891669892007
3.77843445900362
3.740818833000958
3.749061916998471
  Time (mean ± σ):      3.966 s ±  0.027 s    [User: 3.899 s, System: 0.050 s]
  Range (min … max):    3.944 s …  4.008 s    5 runs

Sqlean

 hyperfine --shell /opt/homebrew/bin/fish  --runs 5 --show-output  'uv run python litecli/performance.py'
Benchmark 1: uv run python litecli/performance.py
3.648521625000285
3.724126499990234
3.765500624998822
3.648744167003315
3.666765707996092
  Time (mean ± σ):      3.921 s ±  0.083 s    [User: 3.824 s, System: 0.066 s]
  Range (min … max):    3.854 s …  4.027 s    5 runs

I don't see much deviation in performance and I think it should be okay.

Let me know what you think.

[amjith]

I am definitely intrigued. I'll give it a shot today.

One thing to consider is the additional friction in installation. The sqlean package is a C-extension. I see they have wheels for a supported platforms. But this does introduce an additional dependency that is not part of the std lib. I'm wondering if we can make this an optional package. For instance, a power user could install pip install litecli[sqlean] which adds the extra dependency and we integrate it into our code base as

try:
    import sqlean
except ImportError:
    sqlean = None

Then check for sqlean being not None before we use it.

Just some initial thoughts.

[kracekumar]

But this does introduce an additional dependency that is not part of the std lib. I'm wondering if we can make this an optional package. For instance, a power user could install pip install litecli[sqlean]

That's an another approach. We can update docs to say it. That also means majority of the users have to discover the feature from reading the docs.

try:
    import sqlean
except ImportError:
    sqlean = None

Above code we can look for sqlean if not fallback to sqlite3

try:
    import sqlean as sqlite3
except ImportError:
    import sqlite3

And also fix import exceptions in the code. This can be downside. I see everytime any import from sqlite3 module needs to always have try/except block to support the same imports from both the library.

[kracekumar]

Closing this PR in favour of #221