RR语言避坑指南

Vectorization

• - Loops are slow — use apply(), lapply(), sapply(), or INLINECODE3
• Vectorized functions operate on whole vectors — sum(x) not INLINECODE5
• INLINECODE6 is vectorized — if is not, use ifelse() for vector conditions
• Column operations faster than row — R is column-major

Indexing Gotchas

• - R is 1-indexed — first element is x[1], not INLINECODE10
• INLINECODE11 returns empty vector — not error, silent bug
• Negative index excludes — x[-1] removes first element
• INLINECODE13 extracts single element — [ returns subset (list stays list)
• INLINECODE15 drops to vector — use df[, 1, drop = FALSE] to keep data frame

NA Handling

• - NA propagates — 1 + NA is NA, NA == NA is INLINECODE20
• Use is.na() to check — not INLINECODE22
• Most functions need na.rm = TRUE — mean(x) returns NA if any NA present
• INLINECODE25 removes rows with any NA — may lose data unexpectedly
• INLINECODE26 returns logical vector — rows without NA

Factor Traps

• - Old R converted strings to factors by default — use stringsAsFactors = FALSE or modern R
• INLINECODE28 shows categories — but factor values are integers internally
• Adding new value not in levels gives NA — use INLINECODE29
• INLINECODE30 gives level indices — use as.numeric(as.character(factor)) for values
• Dropping unused levels: droplevels() — or factor() again

Recycling

• - Shorter vector recycled to match longer — c(1,2,3) + c(10,20) gives INLINECODE35
• No error if lengths aren't multiples — just warning, easy to miss
• Single values recycle intentionally — x + 1 adds 1 to all elements

Data Frames vs Tibbles

• - Tibble never converts strings to factors — safer defaults
• Tibble never drops dimensions — df[, 1] stays tibble
• Tibble prints better — shows type, doesn't flood console
• INLINECODE38 to convert — from tibble or dplyr package

Assignment

• - <- is idiomatic R — = works but avoided in style guides
• INLINECODE43 assigns to parent environment — global assignment, usually a mistake
• INLINECODE44 right assignment exists — rarely used, confusing

Scope

• - Functions look up in parent environment — can accidentally use global variable
• Local variable shadows global — same name hides outer variable
• INLINECODE45 creates isolated scope — variables don't leak out

Common Mistakes

• - T and F can be overwritten — use TRUE and FALSE always
• INLINECODE50 fails on empty x — gives c(1, 0), use INLINECODE52
• INLINECODE53 vs sample(c(5)) — different! first gives 1:5 permutation
• String splitting: strsplit() returns list — even for single string