library(tidyverse)
library(dplyr)
library(tidyr)
dsl_csv <-"~/dsl/121625_DSL_frame.csv"
dsl_df <- read.csv(dsl_csv, header = TRUE) 
#4.1
weekly_summary <- dsl_df |>
  group_by(week_index, source, isAuthorWMF)|>
  summarise(
    tasks_made = sum(!is.na(resolution_outcome)),
    count_resolution_outcome = sum(dsl_score),
    author_closer_sum  = sum(author_closer == TRUE),
    median_olmo_EP_prop_adac = median(olmo_EP_prop_adac),
    median_olmo_TSOL_prop_adac = median(olmo_TSOL_prop_adac),
    median_olmo_RK_prop_adac = median(olmo_RK_prop_adac),
    median_comments_before_resolution = median(n_comments_before)
  ) |>
  mutate(isAuthorWMF = factor(isAuthorWMF, levels = c("FALSE", "BzImport", "TRUE")))

tasks_created <- ggplot(
  weekly_summary,
  aes(
    x=week_index,
    y=tasks_made, 
    fill=isAuthorWMF
  )
) + 
  facet_grid(source ~ ., 
             scales = "free_y",
             labeller = labeller(source = c("c1" = "VisualEditor", 
                                            "c2" = "HTTPS-login", 
                                            "c3" = "HTTP-deprecation"))) +  
  geom_col(position = position_dodge(width = 0.9), width = 0.8) +
  geom_vline(data = weekly_summary |> filter(source == "c1"), 
             aes(xintercept = -29), 
             linetype = "dotted", color = "black", linewidth = 0.5) +
  geom_vline(data = weekly_summary |> filter(source == "c1"), 
             aes(xintercept = -9), 
             linetype = "dotted", color = "black", linewidth = 0.5) +
  geom_vline(data = weekly_summary |> filter(source == "c1"), 
             aes(xintercept = -4), 
             linetype = "3313", color = "black", linewidth = 0.5) +
  geom_vline(data = weekly_summary |> filter(source == "c2"), 
             aes(xintercept = -99), 
             linetype = "dotted", color = "black", linewidth = 0.5) +
  geom_vline(data = weekly_summary |> filter(source == "c2"), 
             aes(xintercept = -4),   
             linetype = "3313", color = "black", linewidth = 0.5) +
  geom_vline(data = weekly_summary |> filter(source == "c3"), 
             aes(xintercept = -97), 
             linetype = "dotted", color = "black", linewidth = 0.5) +
  geom_vline(data = weekly_summary |> filter(source == "c3"), 
             aes(xintercept = -3), 
             linetype = "3313", color = "black", linewidth = 0.5) +
  geom_vline(xintercept = 0, linetype = "dashed", color = "black", linewidth = 0.5) +
  geom_text(
    data = subset(weekly_summary, source == "c1" & week_index ==10),
    aes(x=week_index, y=120, label='Opt-out deployment'),
    size = 3) + 
  geom_text(
    data = subset(weekly_summary, source == "c1" & week_index == -21),
    aes(x=week_index, y=120, label='Opt-in Testing'),
    size = 3) + 
  geom_text(
    data = subset(weekly_summary, source == "c2" & week_index == -18),
    aes(x=week_index, y=20, label='Deployment Announcement'),
    size = 3) + 
  theme_minimal() + 
  scale_fill_viridis_d(
    breaks = c("FALSE", "TRUE", "BzImport"),    
    labels = c("External Contributor", "WMF-affiliate", "BzImport")  
  ) + 
  labs(
    x = "Weeks from Feature Deployment", 
    y = "Count of Tasks Created", 
    fill = "Task Author"
  ) + 
  theme(legend.position = "top")
tasks_created
ggsave(
  filename = "011925_tasks_created.png",
  plot = tasks_created,
  width = 8,    # inches
  height = 4,   # inches
  dpi = 800     # high resolution
)


#4.2 plot comparing the TTR for different things 
ttr_trajectory <- dsl_df |>
  mutate(ttr_weeks = TTR_hours / 168) |>
  mutate(isTriaged = if_else(priority == 'Needs Triage',
                             "Not Triaged",
                             "Triaged")) |>
  group_by(week_index, isTriaged, source) |>
  summarise(
    count = n(),
    mean_ttr = mean(ttr_weeks, na.rm = TRUE),
    sd_ttr = sd(ttr_weeks, na.rm = TRUE)
  )
ttr_trajectory_plot <- ttr_trajectory |>
  filter(week_index >= -13) |>
  filter(isTriaged == "Not Triaged") |>
  ggplot(aes(x = week_index)) +
  # Line for mean TTR
  geom_line(aes(y = mean_ttr, color = "Mean TTR"), linewidth = 1) +
  # Ribbon for standard deviation
  geom_ribbon(aes(ymin = mean_ttr - sd_ttr, ymax = mean_ttr + sd_ttr),
              fill = "lightblue", alpha = 0.4) +
  # Line for count of tasks
  geom_point(aes(y = count, 
                color = "Count of New Tasks"), linewidth = 1, linetype = "dashed") +
  # Facet the plot by source and triaged status
  facet_wrap(source ~ isTriaged, scales = "free_y") +
  labs(
    title = "TTR by Source and Triage Status (TODO)",
    x = "Week Index",
    y = "Mean TTR (in weeks)",
    color = "Metrics"
  ) +
  scale_color_manual(values = c("Mean TTR" = "blue", "Count of New Tasks" = "red")) +
  theme_minimal() 
ttr_trajectory_plot


ttr_boxplot <- dsl_df |>
  filter(priority == "Needs Triage" | 
           priority == "Unbreak Now!" |
           priority == "High") |>
  filter(week_index >= -13) |>
  ggplot(
    aes(
      x=as.factor(week_index),
      y= TTR_hours/168, 
      color=priority,
    )
  ) + 
  facet_grid(source ~ ., 
             scales = "free_y",
             labeller = labeller(source = c("c1" = "VisualEditor", 
                                            "c2" = "HTTPS-login", 
                                            "c3" = "HTTP-deprecation"))) + 
  geom_boxplot(outlier.shape = NA) +
  theme_minimal() + 
  coord_cartesian(ylim = c(0, 112)) + 
  geom_text(
    data = subset(dsl_df |>
                    filter(priority == "Needs Triage" | 
                             priority == "Unbreak Now!" |
                             priority == "High"), source == "c1" & week_index == 12),
    aes(x=week_index, y=80, label='Opt-out Deployment'),
    color = "black",
    size = 3) + 
  geom_vline(xintercept =14, linetype = "dashed", color = "black", linewidth = 0.5) +
  scale_color_viridis_d(option='turbo') + 
  labs(x = "Weeks from Release", 
       y = "Time to Resolution (weeks)", 
       color = "Priority Tag") +
  theme(legend.position = "top")
ttr_boxplot
ggsave(
  filename = "011925_ttr_boxplot.png",
  plot = ttr_boxplot,
  width = 8,    # inches
  height = 4,   # inches
  dpi = 800     # high resolution
)
#4.3 plot comparing machine labels of information type 
dsl_df_long <- dsl_df %>%
  filter(isAuthorWMF != "BzImport")|>
  pivot_longer(
    cols = c(olmo_EP_prop_adac, olmo_RK_prop_adac, olmo_TSOL_prop_adac),
    names_to = "tag",
    values_to = "proportion"
  ) %>%
  mutate(tag = gsub("olmo_|_prop_adac", "", tag),
         tag = case_when(
           tag == "EP" ~ "Existent Problem",
           tag == "RK" ~ "Record Keeping",
           tag =="TSOL" ~ "Solutions"
         )) 

olmo_comparison <- ggplot(
  dsl_df_long,
  aes(
    x = tag,
    y = proportion, 
    fill = isAuthorWMF,
  )
) + 
  facet_grid(source ~ ., 
             scales = "free_y",
             labeller = labeller(source = c("c1" = "VisualEditor", 
                                            "c2" = "HTTPS-login", 
                                            "c3" = "HTTP-deprecation"))) +
  geom_boxplot() +
  theme_minimal() + 
  scale_fill_viridis_d(
    breaks = c("FALSE", "TRUE", "BzImport"),    
    labels = c("External Contributor", "WMF-affiliate", "BzImport")  
  ) + 
  labs(
    x = "Issue Information Type Category", 
    y = "% of sentences machine-labeled", 
    color = "Is Author WMF?",
    fill = "Is Author WMF?"
  ) + 
  theme(legend.position = "top")
olmo_comparison
ggsave(
  filename = "011925_machine_label_comparison.png",
  plot = olmo_comparison,
  width = 8,    # inches
  height = 4,   # inches
  dpi = 800     # high resolution
)