explain for a SQL query is affected by options(digits.secs)

[chicotobi]

Expectation: Copy&Paste of the explain output into the SQL CLI should yield the same result as executing the query.
But here the output of explain depends on options(digits.secs). Is that correct?

rm(list=ls())

library(RSQLite)
library(dplyr)
library(dbplyr)

# Create example table in-memory
con <- dbConnect(RSQLite::SQLite(), ":memory:", extended_types = TRUE)
con %>% dbExecute("CREATE TABLE tmp (t datetime );")
time1 <- "2024-01-11 09:17:12.345" %>% as.POSIXct()
df <- data.frame(t=time1)
con %>% dbAppendTable("tmp",df)

# Query with a datetime filter - digits.secs is 0, result is not affected
time2 <- "2024-01-11 09:17:12.678" %>% as.POSIXct()
options(digits.secs=0)
con %>% tbl("tmp") %>% filter(t<=time2) %>% explain()
con %>% tbl("tmp") %>% filter(t<=time2) %>% collect()

# Query with a datetime filter - digits.secs is 3
options(digits.secs=3)
con %>% tbl("tmp") %>% filter(t<=time2) %>% explain()
con %>% tbl("tmp") %>% filter(t<=time2) %>% collect()

Output:

<SQL>
SELECT *
FROM `tmp`
WHERE (`t` <= '2024-01-11T08:17:12Z')

<PLAN>
  id parent notused   detail
1  2      0       0 SCAN tmp
<SQL>
SELECT *
FROM `tmp`
WHERE (`t` <= '2024-01-11T08:17:12.677Z')

<PLAN>
  id parent notused   detail
1  2      0       0 SCAN tmp

[hadley]

That's coming from this code: https://github.com/tidyverse/dbplyr/blob/main/R/db-escape.R#L49-L53

So it seems probably fine to me? I don't love that a global option affects this, but that option is baked fairly deeply in to R itself, and if that option didn't exist, we'd have to find some other way to pass the number of digits.