1
0
mirror of https://github.com/jbranchaud/til synced 2026-01-03 15:18:01 +00:00

Add Unable To Infer Data Type In Production as a Postgres TIL

This commit is contained in:
jbranchaud
2025-01-28 18:40:09 -06:00
parent 46ad33df7e
commit 61fc021f52
2 changed files with 47 additions and 1 deletions

View File

@@ -10,7 +10,7 @@ pairing with smart people at Hashrocket.
For a steady stream of TILs, [sign up for my newsletter](https://crafty-builder-6996.ck.page/e169c61186).
_1578 TILs and counting..._
_1579 TILs and counting..._
See some of the other learning resources I work on:
- [Ruby Operator Lookup](https://www.visualmode.dev/ruby-operators)
@@ -905,6 +905,7 @@ See some of the other learning resources I work on:
- [Two Ways To Compute Factorial](postgres/two-ways-to-compute-factorial.md)
- [Two Ways To Escape A Quote In A String](postgres/two-ways-to-escape-a-quote-in-a-string.md)
- [Types By Category](postgres/types-by-category.md)
- [Unable To Infer Data Type In Production](postgres/unable-to-infer-data-type-in-production.md)
- [Union All Rows Including Duplicates](postgres/union-all-rows-including-duplicates.md)
- [Use A psqlrc File For Common Settings](postgres/use-a-psqlrc-file-for-common-settings.md)
- [Use A Trigger To Mirror Inserts To Another Table](postgres/use-a-trigger-to-mirror-inserts-to-another-table.md)

View File

@@ -0,0 +1,45 @@
# Unable To Infer Data Type In Production
Inspired by [You Probably Don't Need Query
Builders](https://mattrighetti.com/2025/01/20/you-dont-need-sql-builders), I
wrote a query in one of my applications that has filter clauses that get
short-circuited if the filter value hasn't been included.
That query looked something like this:
```ruby
@tags =
Tag.where("? is null or normalized_value ilike ?", normalized_query, "%#{normalized_query}%")
.order(:normalized_value)
.limit(10)
```
The `normalized_value ilike ?` filtering won't be applied if the
`normalized_query` value isn't present (`nil`). This helps me avoid writing
messy ternaries or if-else conditional query building madness.
Unfortunately, when I shipped this query to production, the page started
failing and Postgres was reporting this error in the logs.
```
Caused by: PG::IndeterminateDatatype (ERROR: could not determine data type of parameter $1)
```
The query is prepared as a parameterized statement and Postgres appears to be
unable to determine the datatype of the first parameter (`$1`) —
`normalized_query`.
I was unable to reproduce the issue in development. It was only occuring in
production. Until I can come up with a root cause analysis, I have the
following fix that does a casting to `text`. This helps out with the type
inference and makes the issue go away.
```ruby
@tags =
Tag.where("cast(? as text) is null or normalized_value ilike ?", normalized_query, "%#{normalized_query}%")
.order(:normalized_value)
.limit(10)
```
Interestingly, this person using `pgtyped` [ran into the exact same issue with
the same type of query](https://github.com/adelsz/pgtyped/issues/354).