Skip to content

⚡ Bolt: 데이터프레임 서브셋팅 할당 최적화로 성능 개선 #153

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
seonghobae wants to merge 1 commit into from
Closed

⚡ Bolt: 데이터프레임 서브셋팅 할당 최적화로 성능 개선 #153

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions .jules/bolt.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,3 +16,6 @@
## 2025-02-12 - R 언어에서 반복적인 mirt 모델 생성 시 불필요한 데이터프레임 부분집합 추출 최적화
**Learning:** R에서 데이터프레임의 특정 열을 추출하는 작업(`df[cols]`)은 O(N)의 메모리 복사를 수반합니다. `autoFIPC`에서 `mirt` 모델의 파라미터를 설정하거나 호출하는 과정 중에 `newformXDataK[colnames(newFormModel@Data$data)]` 코드가 반복해서 사용되었고, 심지어 `ncol()`을 위해 단순히 개수를 구할 때도 사용되어 불필요한 메모리 할당과 오버헤드를 초래했습니다.
**Action:** 조건문이나 반복문 내부에서 불필요하게 데이터프레임 부분집합 연산이 반복되지 않도록 외부에서 한 번만 `linkedFormData <- newformXDataK[colnames(newFormModel@Data$data)]`로 캐싱(caching)한 뒤, `ncol(linkedFormData)`와 `data = linkedFormData` 형태로 재사용하여 메모리 복사와 O(N) 오버헤드를 방지해야 합니다.
## 2024-07-19 - R Dataframe Dispatch Overhead Optimization
**Learning:** In R, updating a specific column based on a conditional match is significantly faster using direct vector subsetting (e.g., `df$col[df$idx == 'val'] <- new_val`) compared to two-dimensional subsetting (e.g., `df[df$idx == 'val', 'col'] <- new_val`). Direct vector assignment bypasses the `[<-.data.frame` method dispatch overhead (which checks dimensions, factor levels, and often deeply copies the data frame) in favor of O(1) list access and C-level vector modification.
**Action:** Always prefer 1D vector subsetting for assignments in R dataframes to avoid performance degradation from method dispatch and memory copies, especially inside iterative processes or large data updates.
51 changes: 28 additions & 23 deletions R/aFIPC.R
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -598,15 +598,16 @@ autoFIPC <-
# Preserve mirt's structural estimability flags. Forcing every row TRUE
# frees boundary parameters such as 2PL g/u and makes the Hessian unstable.

NewScaleParms[NewScaleParms$item == 'GROUP', "est"] <- FALSE
OldScaleParms[OldScaleParms$item == 'GROUP', "est"] <- FALSE
# ⚡ Bolt: Use direct vector subsetting (O(1)) instead of 2D data frame assignment to avoid `[<-.data.frame` dispatch overhead
NewScaleParms$est[NewScaleParms$item == 'GROUP'] <- FALSE
OldScaleParms$est[OldScaleParms$item == 'GROUP'] <- FALSE

NewScaleParms[NewScaleParms$name == "COV_11", "est"] <- TRUE
OldScaleParms[OldScaleParms$name == "COV_11", "est"] <- TRUE
NewScaleParms$est[NewScaleParms$name == "COV_11"] <- TRUE
OldScaleParms$est[OldScaleParms$name == "COV_11"] <- TRUE

if (itemtype == 'Rasch') {
NewScaleParms[NewScaleParms$name == "a1", "est"] <- FALSE
OldScaleParms[OldScaleParms$name == "a1", "est"] <- FALSE
NewScaleParms$est[NewScaleParms$name == "a1"] <- FALSE
OldScaleParms$est[OldScaleParms$name == "a1"] <- FALSE
}

#IPD
Expand Down Expand Up @@ -641,9 +642,10 @@ autoFIPC <-
colnames(IPDData) <- paste0('X', 1:IPDItemCount)
print(IPDItemNamesOldForm)
print(IPDItemNamesNewForm)
# ⚡ Bolt: Fix range evaluation for row assignments
IPDData[1:nrow(oldformYDataK), ] <-
oldformYDataK[, IPDItemNamesOldForm]
IPDData[nrow(oldformYDataK) + 1:nrow(newformXDataK), ] <-
IPDData[(nrow(oldformYDataK) + 1):(nrow(oldformYDataK) + nrow(newformXDataK)), ] <-
newformXDataK[, IPDItemNamesNewForm]

# IPD estimation
Expand Down Expand Up @@ -789,11 +791,12 @@ autoFIPC <-
message(' Newform Parms: ', paste(NewScaleParms[newIdx, "value"], collapse = ' '))
message(' Oldform Parms: ', paste(OldScaleParms[oldIdx, "value"], collapse = ' '))

NewScaleParms[newIdx, "value"] <-
OldScaleParms[oldIdx, "value"]
message(' Linkedform Parms: ', paste(NewScaleParms[newIdx, "value"], collapse = ' '), '\n')
# ⚡ Bolt: Direct vector subsetting
NewScaleParms$value[newIdx] <-
OldScaleParms$value[oldIdx]
message(' Linkedform Parms: ', paste(NewScaleParms$value[newIdx], collapse = ' '), '\n')

NewScaleParms[newIdx, "est"] <-
NewScaleParms$est[newIdx] <-
FALSE
} else {
message(
Expand All @@ -813,17 +816,18 @@ autoFIPC <-
newBetaIdx <- NewScaleParms$item == 'BETA'
oldBetaIdx <- OldScaleParms$item == 'BETA'

NewScaleParms[newBetaIdx, "value"] <-
OldScaleParms[oldBetaIdx, "value"]
NewScaleParms[newBetaIdx, "est"] <-
# ⚡ Bolt: Direct vector subsetting
NewScaleParms$value[newBetaIdx] <-
OldScaleParms$value[oldBetaIdx]
NewScaleParms$est[newBetaIdx] <-
FALSE

message('applying BETA parameter as linking')

message(
' Linkedform Parms: ',
paste0(
NewScaleParms[newBetaIdx, "value"],
NewScaleParms$value[newBetaIdx],
' '
),
'\n'
Expand Down Expand Up @@ -858,13 +862,14 @@ autoFIPC <-
new_mean11_idx <- NewScaleParms$name == "MEAN_11"
old_mean11_idx <- OldScaleParms$name == "MEAN_11"

NewScaleParms[new_cov11_idx, "est"] <- FALSE
OldScaleParms[old_cov11_idx, "est"] <- FALSE
NewScaleParms[new_mean11_idx, "est"] <- FALSE
OldScaleParms[old_mean11_idx, "est"] <- FALSE
# ⚡ Bolt: Direct vector subsetting
NewScaleParms$est[new_cov11_idx] <- FALSE
OldScaleParms$est[old_cov11_idx] <- FALSE
NewScaleParms$est[new_mean11_idx] <- FALSE
OldScaleParms$est[old_mean11_idx] <- FALSE

NewScaleParms[new_cov11_idx, "value"] <- 1
OldScaleParms[old_mean11_idx, "value"] <- 0
NewScaleParms$value[new_cov11_idx] <- 1
OldScaleParms$value[old_mean11_idx] <- 0
}
if (freeMEAN == T) {
LinkedModelSyntax <-
Expand All @@ -875,8 +880,8 @@ autoFIPC <-
'MEAN = F1'
))

NewScaleParms[NewScaleParms$name == "MEAN_1", "est"] <- TRUE
OldScaleParms[OldScaleParms$name == "MEAN_1", "est"] <- TRUE
NewScaleParms$est[NewScaleParms$name == "MEAN_1"] <- TRUE
OldScaleParms$est[OldScaleParms$name == "MEAN_1"] <- TRUE
} else {
LinkedModelSyntax <-
mirt::mirt.model(paste0(
Expand Down
Loading