#community-help

Resolving Floating Point Arithmetic in Facets

TLDR Claudiu reported an issue with floating point value representation in facets and found a workaround by converting everything to int32. Kishore Nallan validated this fix and also committed to finding a solution. Later, Erick faced a related error.

21mo

Join the chat

May 19, 2021 (31 months ago)

Photo of md5-d619094f9c50a3e0a4927dc0e63ba30d

Claudiu

07:40 PM

does anybody know about this issue?

Photo of md5-8813087cccc512313602b6d9f9ece19f

Jason

08:18 PM

Ah good ol' floating point arithmetic. Could you open a github issue about this Claudiu?

Claudiu

08:22 PM

hey, i will tomorrow, in the document the value looks ok in the hits, just on the facets value is the problem

May 21, 2021 (30 months ago)

Photo of md5-4e872368b2b2668460205b409e95c2ea

Kishore Nallan

10:03 AM

Claudiu I'm not able to reproduce this issue. I indexed a single document with a facet field value of 113.4 and facet on the same field during searching. See the screenshot: the response looks fine.

10:03

Kishore Nallan

10:03 AM

If you can share a small snippet that reproduces this issue, that will help us narrow down the issue.

Claudiu

12:09 PM

hey Kishore Nallan but in the document the value of the field "points" is 113.4 and in the facet is returned as 113.400002. Why is this happening?

12:11

Claudiu

12:11 PM

Kishore Nallan

12:16 PM

Claudiu That's because many floating point numbers cannot be represented exactly, so when they are serialized they might have some trailing decimal places at the end. In general, for facets it is better to store price in cents (as integer) instead of a float.

12:17

Kishore Nallan

12:17 PM

In your original post though, you said that 113.4 is changed to 114.300003 -> such a large difference should not happen. Was that a typo?

Claudiu

12:31 PM

look at the CB field 63.3 is returned in facet like 63.299999

12:33

Claudiu

12:33 PM

the thing is i have to do a filter in the sidebar, and i need the exact value

12:33

Claudiu

12:33 PM

because i will filter by that value

12:33

Claudiu

12:33 PM

a workaround is to convert everything to int32

Kishore Nallan

12:33 PM

Yes, it's the same issue. When you convert a number from a floating point representation to string, this can happen unless we know how many decimal places the original precision of the float was. In this case, the best way to work around this issue is to store the float as an integer in cents.

Claudiu

12:34 PM

thanks

12:34

Claudiu

12:34 PM

like i guessed

Kishore Nallan

12:35 PM

In the mean time, I will see if there is anything we can do here. It's a notoriously difficult problem because of the way a float is represented.

Claudiu

12:36 PM

i understand

12:36

Claudiu

12:36 PM

i will user. integer

Kishore Nallan

12:36 PM

👍

Claudiu

12:37 PM

you are doing a good job

Kishore Nallan

01:28 PM

Claudiu Found a way to fix this. I can cut you a RC build where this is addressed, if you would like.

Claudiu

01:28 PM

01:28

Claudiu

01:28 PM

i already used integer

01:28

Claudiu

01:28 PM

but i can test it

Kishore Nallan

01:28 PM

No worries, then! Just wanted to check 🙂

Claudiu

01:29 PM

you guys are so fast 😄

Kishore Nallan

01:29 PM

😄

Claudiu

01:29 PM

i hope this community is getting bigger and bigger

Feb 01, 2022 (21 months ago)

Photo of md5-5d002d197dc556e6bc619deaa3a8aba7

Erick

07:06 PM

hey Kishore Nallan, I'm getting the error 400: {"message": "Field cost should be a string or a string array."} . The actual value is float. What do you think is the issue here?

Typesense

Lightning-fast, open source search engine for everyone | Knowledge Base powered by Struct.AI

Indexed 2786 threads (79% resolved)

Join Our Community Manage subscription

Similar Threads

Resolving Float Value Issues in PHP and Typesense.

Zildjian was having issues with float values being treated as int64 in Typesense. Kishore Nallan identified and rectified the issue, guiding the user through a software upgrade process.

2mo

Setting dynamic range faceting in Typesense

Jacob requested dynamic range faceting in Typesense, and explanation for fetching max/min values. Kishore Nallan clarified pricing bucket sizes were user dependent and explained the stats return as part of facet results.

1mo

Issue with Dynamic Nested Field Type in TS Collection Schema

Justin encountered an issue with dynamic nested fields not showing as `float`. Jason suggested reproducing the error with a snippet, while Kishore Nallan was tagged for assistance.

6mo

Quick Retrieval of Min, Max, and Median Values from Search Query

Prajjval sought a more efficient way to find the min, max, and median values from a search query result. Jason suggested faceting. However, Prajjval clarified that faceting prolonged document upload due to a data type error, which was later resolved.

31mo

Precision Issues with Int64 Numeric Field in Go Client

SamHendley expressed concerns about precision and rounding problems with an int64 numeric field using a timestamp. The problem was traced back to a conversion issue in the go client, prompting SamHendley to suggest updating the openapi spec to use double for reduced errors.

11mo