Does it support fuzzy search on multiple and related fields?

[Tragio]

Hi @mbi 👋

First, thank you for your incredible work.

While exploring search solutions that don't rely on Meilisearch, Typesense, or similar external dependencies, I came across your integration and found it very promising. I have a couple of questions 🤔

1. Does the fuzzy search work with related fields? For example, would it be able to search within user__addresses.street?

from django.contrib.auth.models import User
from django.db import models

class Address(models.Model):
    user = models.ForeignKey(User, on_delete=models.CASCADE, related_name='addresses')
    street = models.CharField(max_length=255)
    city = models.CharField(max_length=100)
    state = models.CharField(max_length=100)
    country = models.CharField(max_length=100)
    zip_code = models.CharField(max_length=20)

2. Can multiple fields be configured for fuzzy search simultaneously?

Thank you again for your time and effort on this project. 🚀

[mbi]

Hello. Short answer to both questions: I don't know.

To be honest I've started this app more as a test of the feasibility of integrating pg_search into Django, but the experience hasn't been very satisfying so far:

• ParadeDB looks very powerful on paper, but I haven't been able to match the performance of ts_vector (see Add pg_search support to Django paradedb/paradedb#1098)
• pg_search's custom syntax is complex and, probably due to my lack of deep understanding of the internals of django.db, integrating it into Django has been lacking.
• Most of the times I've tried upgrading ParadeDB, things have randomly broken

In a nutshell, this repo is little more than a proof of concept and should probably not be used for anything more than that.

I'd love to expand the pg_search integration and improve this app, but at the moment my opinion is that it's probably easier to use ts_vector or a dedicated search engine (Meili, ElasticSearch) rather than going with pg_search.

[mbi]

Hello @Tragio an update on this:

1. Yes: as of today, fuzzy search works on related fields.
2. No: multiple fields can't be configured yet, but I expect to add this in a near future.

Also, I've updated my benchmark management command to span a much larger dataset (approx 800'000 items) and in this scenario ParadeDB is much more performant than Django's native ts_vector.

[mbi]

@Tragio it just occurred to me, the second use-case, i.e. matching multiple fields can probably easily be achieved by chaining two filters, i.e.

Book.objects.filter(description__fuzzy_term_search='woooden').filter(title__fuzzy_term_search='woooden').annotate(score=Score('title')).order_by('-score')

Or am I missing something?

[philippemnoel]

Hello @Tragio an update on this:

1. Yes: as of today, fuzzy search works on related fields.
2. No: multiple fields can't be configured yet, but I expect to add this in a near future.

Also, I've updated my benchmark management command to span a much larger dataset (approx 800'000 items) and in this scenario ParadeDB is much more performant than Django's native ts_vector.

Just wanted to pop by and say yay! 🎉 I'm one of the maintainers of pg_search. We've:
1- Improved performance a lot
2- We really expect ParadeDB to be used on much larger scale than ts_vector, and see the performance benefits of ParadeDB get better and better compared to ts_vector at scale.

Excited to see this project grow, and let us know if/how we can help!

[Tragio]

Hi @mbi and @philippemnoel 👋

Thank you very much for the replies and so sorry for the late reply, was on vacation eheh 🏖

At the moment, I'm using Meilisearch as it is straightforward to use, however, I need to keep keys and data in sync. That is why I was trying other solutions. I can't recall the details, but neither ts_vector nor pg_trgm was doing the fuzzy search or supporting/providing relevant results with multiple search fields across models.

I also encountered other people in the same boat. https://forum.djangoproject.com/t/replacing-algolia-meilisearch-typesense-with-postgres/30356

I still have to find out if parade_db is a good fit for my use case 🤔