I've seen this exact bug trip up plenty of developers transitioning from base R to the Tidyverse. You're working with what looks like a standard data frame, you try to pull a single column to run a calculation, and suddenly R is throwing a cryptic error telling you that your numeric data isn't actually numeric.

library(tibble)

# Creating a small dataset of server response times
logs <- tibble(
  endpoint = c("/home", "/api/v1/user", "/login", "/settings"),
  latency = c(120, 450, 80, 210)
)

# I want the average latency, so I'll grab the second column
avg_latency <- mean(logs[, 2])

# Error in mean.default(logs[, 2]) : 
# argument is not numeric or logical: returning NA

The "Vector vs. Tibble" Indexing Trap

If logs were a base R data.frame, that code would work perfectly. In base R, when you use a single bracket [, 2] to select one column, R "simplifies" the result. It says, "Oh, you only want one column? I'll just give you a plain old vector."

Tibbles are different. They are designed to be consistent. A tibble is a data frame, and they believe that slicing a data frame should always return another data frame (or tibble). So, logs[, 2] isn't returning a numeric vector of latencies; it's returning a 4x1 tibble that contains a numeric vector. The mean() function has no idea what to do with a tibble object, so it fails.

Forcing a Vector Return with Double Brackets

To fix this, you need to be explicit about wanting the underlying data, not a miniature table. The cleanest way to do this is using double brackets [[ ]]. While single brackets slice the table, double brackets "reach inside" and grab the actual element.

# This reaches inside the tibble and pulls out the actual numeric vector
avg_latency <- mean(logs[[2]]) 
# Result: 215

# Or, even better, use the column name for readability
avg_latency <- mean(logs[["latency"]])
# Result: 215

I generally recommend using the $ operator (e.g., logs$latency) for quick interactive work, but [[ ]] is more robust when you're writing functions where the column index or name might be stored in a variable.

Why we put up with this (The Tibble Advantage)

You might be wondering why we'd want a "stricter" data frame that breaks base R indexing. In my experience, the trade-off is worth it for two main reasons: printing and type safety.

• Sane Printing: If you accidentally print a base data frame with 100,000 rows, R will try to dump every single one of them into your console, likely freezing your IDE. A tibble only prints the first 10 rows and tells you how many more are hidden. It's a lifesaver.
• No More Secret Conversions: In older versions of base R, data.frame() would automatically convert strings to factors. This caused endless headaches. Tibbles never change the type of your input data. What you put in is exactly what you get out.

📋 Practical Task

Debugging the Monthly Revenue Summary

You have been handed a script that calculates the total revenue from a tibble of sales. However, the original author used base R indexing logic, and the script is currently returning NA instead of a sum.

Your Task: Fix the total_rev calculation so that it correctly extracts the amount column as a vector and sums it up.

library(tibble)

sales_data <- tibble(
  transaction_id = c(101, 102, 103, 104),
  customer = c("Alice", "Bob", "Charlie", "Diana"),
  amount = c(150.00, 200.50, 50.25, 300.00)
)

# FIX THIS LINE:
total_rev <- sum(sales_data[, 3])

print(total_rev)