library(plyr)
library(tidyverse)
library(rdd)


#set wd, read in data
try(setwd(dirname(rstudioapi::getActiveDocumentContext()$path)))
readme_df <- read_csv("../final_data/deb_readme_did.csv")
contributing_df <- read_csv("../final_data/deb_contrib_did.csv")
full_df <- read_csv("../final_data/deb_full_data.csv")




#preprocessing for readme_df
colnames(contributing_df) <- c("upstream_vcs_link", "event_date", "event_hash", "before_all_ct", "before_mrg_ct", "after_all_ct", "after_mrg_ct", "before_auth_new", "after_commit_new", "after_auth_new", "before_commit_new")
col_order <- c("upstream_vcs_link", "event_date", "event_hash", "before_all_ct", "after_all_ct", "before_mrg_ct", "after_mrg_ct", "before_auth_new", "after_auth_new", "before_commit_new",  "after_commit_new")

ages <- c()
projects <- c()
for (i in 1:nrow(contributing_df)){
  link <- contributing_df[i,]$upstream_vcs_link
  age <- full_df$age_of_project[full_df$upstream_vcs_link == link]
  project <- full_df$project_name[full_df$upstream_vcs_link == link]
  ages <- c(ages, age)
  if (length(project) != 1){
    project
    break
  } else {
    projects <- c(projects, project)
  }
}
contributing_df$age_of_project = ages
write.csv(contributing_df, "deb_contributing_data_4_19.csv", row.names=FALSE)
readme_df <- readme_df[,col_order]
readme_df$ct_before_all <- str_split(gsub("[][]","", readme_df$before_all_ct), ", ")
readme_df$ct_after_all <- str_split(gsub("[][]","", readme_df$after_all_ct), ", ")
readme_df$ct_before_mrg <- str_split(gsub("[][]","", readme_df$before_mrg_ct), ", ")
readme_df$ct_after_mrg <- str_split(gsub("[][]","", readme_df$after_mrg_ct), ", ")
drop <- c("before_all_ct", "before_mrg_ct", "after_all_ct", "after_mrg_ct")
readme_df = readme_df[,!(names(readme_df) %in% drop)]

#preprocessing for contributing_df


# test <- readme_df$cnt_before_all
# as.numeric(unlist(test[1]))
# test_two <- c()
# iterator <- 0
# for (entry in test) {
#   readme_df$cnt_before_all[iterator] <- as.numeric(unlist(entry))
#   print(as.numeric(unlist(entry)))
#   iterator <- iterator + 1
# }
# test_two
#Yes, need to expand the dataframe, but again, for the sake of clarity, do not want to until analysis step
# https://rpubs.com/phle/r_tutorial_regression_discontinuity_design
## https://www.rdocumentation.org/packages/lme4/versions/1.1-35.2/topics/lmer
new_test <- readme_df[9,]
get_optimal_window <- function(project_row) {
  longer <- project_row |> 
    pivot_longer(cols = starts_with("ct"),
                 names_to = "window",
                 values_to = "count") |>
    unnest(count) 
  longer$observation_type <- gsub("^.*_", "", longer$window)
  longer <- ddply(longer, "observation_type", transform, week=seq(from=0, by=1, length.out=length(observation_type)))
  longer$count <- as.numeric(longer$count)
  #this below line makes the code specific to the all-commits data
  longer <- longer[which(longer$observation_type == "all"),]
  result <- tryCatch({
    #Imbens-Kalyanaraman Optimal Bandwidth Calculation
    optimal_bandwidth <- IKbandwidth(longer$week, longer$count, cutpoint = 26, verbose = FALSE, kernel = "triangular")
    return(optimal_bandwidth)
  }, error = function(e){
    #have tested it with multiple different error-values and all medians/means still hover around 8
    return(8)
  })
}

#this just gets the optimal bandwith window for each project and then appends to lists 
bandwidths <- c()
for (i in 1:nrow(readme_df)){
  bandwidth <- get_optimal_window(readme_df[i,])
  bandwidths <- c(bandwidths, bandwidth)
}
mean(bandwidths) 
#8.574233
median(bandwidths) #8.363088
table(bandwidths)
#from this, I think setting the bandwidth to 8 weeks, two months, the floor 
# of both the median and mean calculations 


#longer <- longer %>%
#  filter(week >= (26 - window_num) & week <= (26 + window_num))
#sapply(longer, class)
#longer$biweekly <- ceiling(longer$week / 2)
#longer <- longer %>%
#  group_by(window, biweekly, observation_type) %>%
#  summarise(biweekly_count = sum(count, na.rm = TRUE))

#testing out analysis below
longer[which(longer$observation_type == "all"),] |> 
  ggplot(aes(x = week, y = count)) + 
  geom_point() +
  geom_vline(xintercept = 26) 

longer[which(longer$observation_type == "all"),] |>
  mutate(D = ifelse(week >= 26, 1, 0)) |>
  lm(formula = count ~ D * I(week - 26)) |> 
  summary()

longer[which(longer$observation_type == "all"),] |>
  select(count, week) |>
  mutate(D = as.factor(ifelse(week >= 26, 1, 0))) |>
  ggplot(aes(x = week, y = count, color = D)) +
  geom_point() + 
  geom_smooth(se = TRUE) + 
  geom_vline(xintercept = 26)