如何用列表条件程序化过滤data.table？求类似pandas query的方法

Great question! Ditching the error-prone eval(parse()) approach for a cleaner, programmatic way to filter data.table makes total sense—let's look at a few approaches that mimic the flexibility of pandas' query() with keyword-style arguments.

1. Direct List Matching (Simplest for Equality Checks) #

data.table has native support for filtering with named lists, which is perfect for your use case where you're checking equality against column values. This avoids string parsing entirely:

library(data.table)
setDT(dat4)

# Your condition list
ab <- list(Clinic = TRUE, Hospital = FALSE)

# Filter directly using the list + `on` argument
ad <- dat4[ab, on = names(ab)]

How this works: The on parameter tells data.table which columns to match against the list's keys. It automatically translates the list into the equivalent of Clinic == TRUE & Hospital == FALSE under the hood.

2. Dynamic Expression Building (For More Complex Conditions) #

If you need to handle more than just equality checks (e.g., >, <, %in%), you can build a filter expression programmatically using rlang tools—no string parsing required:

library(data.table)
library(rlang)
library(purrr)

setDT(dat4)
ab <- list(Clinic = TRUE, Hospital = FALSE)

# Build a list of conditions (e.g., Clinic == TRUE, Hospital == FALSE)
condition_list <- imap(ab, ~ sym(.y) == .x)

# Unpack the conditions into a single expression joined by `&`
filter_expr <- expr(!!!condition_list)

# Apply the filter
ad <- dat4[eval(filter_expr)]

This approach is safer and more maintainable than eval(parse()) because it works directly with R's expression objects instead of raw strings. You can easily modify it for other operators—just swap == with > or %in% as needed.

3. Custom Wrapper Function (Mimic pandas' query() Syntax) #

If you want a function that feels exactly like pandas' query() (accepting keyword arguments directly), wrap the logic above into a reusable function:

library(data.table)
library(rlang)
library(purrr)

dt_query <- function(dt, ...) {
  # Capture keyword arguments as a list
  filter_args <- list(...)
  
  # Build and evaluate the filter expression
  filter_expr <- expr(!!!imap(filter_args, ~ sym(.y) == .x))
  dt[eval(filter_expr)]
}

# Use it just like pandas' query()!
ad <- dt_query(dat4, Clinic = TRUE, Hospital = FALSE)

Why This Is Better Than Your Original Approach #

Your original eval(parse()) method works, but it has downsides:

• Risky: If column names contain special characters or user input is untrusted, it can lead to unexpected behavior or code injection.
• Hard to debug: String-based conditions are harder to inspect and modify than expression objects.
• Less idiomatic: The methods above align with data.table's native design, making your code more readable for other data.table users.

内容的提问来源于stack exchange，提问作者JeanVuda