phila_whole <- st_union(wards)
bb_phila <- st_bbox(phila_whole, 1.2)

ggplot(phila_whole) +
  geom_sf(fill=strong_green, color=NA, alpha=0.4) +
  geom_sf(
    data=st_crop(sts, bb_phila),
    fill=strong_green, color = "white", size = 1, alpha=0.3
  ) +
  geom_sf(
    data=sts %>% filter(SLDUST == "001"),
    fill=strong_green, color = "white", size = 1
  ) +
  geom_sf_text(
    data = st_crop(sts, bb_phila),
    aes(label=asnum(SLDUST))
  ) +
  ggtitle("State Senate Districts") +
  theme_map_sixtysix() %+replace%
  theme(panel.grid.major = element_blank())

divs$sts <- divs %>%
  st_centroid() %>%
  st_covered_by(sts) %>%
  (function(x){sts$SLDUST[unlist(x)]})

district <- "001"
district_map +
  geom_sf(
    data = wards %>% filter(ward %in% c("02", "08")),
    fill="black",
    alpha = 0.5,
    color=NA
  ) +
  geom_sf(
    aes(alpha = (SLDUST == district)),
    fill="black",
    color = "grey50",
    size=2
  ) +
  geom_sf_text(
    data=wards %>% filter(ward %in% c("02", "08")),
    aes(label = sprintf("Ward %i", asnum(ward))),
    color="white"
  ) +
  scale_alpha_manual(values = c(`TRUE` = 0.2, `FALSE` = 0), guide = FALSE) +
  ggtitle(sprintf("State Senate District %s", asnum(district)))

turnout <- df_major %>%
  filter(is_topline_office) %>%
  group_by(warddiv, year, election) %>%
  summarise(turnout = sum(votes))

maxval <- 20e3
labelmax <- function(x, max) {
  ifelse(
    x == max, 
    paste0(scales::comma(x), "+"), 
    scales::comma(x)
  )
}

district_map +
  geom_sf(
    data=divs %>% 
      left_join(
        turnout %>% filter(year == 2019, election == "general")
      ),
    aes(fill = pmin(turnout/area, maxval)),
    color=NA,
    alpha=0.8
  ) +
  geom_sf(
    data=st_crop(sts, bb_phila), 
    fill=NA, 
    color="white",
    size=2
  ) +
  scale_fill_viridis_c(
    "Votes/\n mile^2", 
    labels=function(x) labelmax(x, maxval)
  ) +
  ggtitle("Votes in the 2019 General")

census <- read_csv("../../data/census/acs_2013_2017_phila_bg_race_income.csv")

maxpop_name <- function(wht, blk, hisp, asn){
  maxes <- apply(cbind(wht, blk, hisp, asn), 1, which.max)
  maxes <- sapply(maxes, function(x) if(length(x) == 1) x else NA)
  c("white", "black", "hispanic", "asian")[maxes]
}

maxpop_val <- function(wht, blk, hisp, asn){
  apply(cbind(wht, blk, hisp, asn), 1, max)
}

census <- census %>%
  mutate(tract = substr(as.character(Geo_FIPS), 1, 11)) %>%
  group_by(tract) %>%
  summarise_at(
    vars(starts_with("pop_")), 
    funs(sum)
  ) %>%
  mutate(pop_total = pop_hisp + pop_nh) %>%
  mutate(
    pwht = pop_nh_white / pop_total,
    pblk = pop_nh_black / pop_total,
    phisp = pop_hisp / pop_total,
    pasn = pop_nh_asian / pop_total
  ) %>%
  mutate(
    maxpop = maxpop_name(pwht, pblk, phisp, pasn),
    maxpop_val = maxpop_val(pwht, pblk, phisp, pasn)
  )

tracts <- st_read("../../data/gis/census/cb_2015_42_tract_500k.shp") %>%
  filter(COUNTYFP == "101") %>%
  st_transform(4326) %>%
  left_join(census, by=c("GEOID" = "tract"))district_map +
  geom_sf(
    data=tracts %>% 
      filter(!is.na(maxpop_val)) %>%
      mutate(maxpop = format_name(maxpop)),
    aes(
      fill=maxpop,
      alpha = 100 * maxpop_val
    ),
    color=NA
  ) +
  geom_sf(
    data=st_crop(sts, bb_phila), 
    fill=NA, 
    color="white",
    size=2
  ) +
  scale_alpha_continuous(
    range=c(0, 1),
    guide=FALSE
  ) +
  scale_fill_manual(
    "Predominant\n Race/Ethn.",
    values = c(
      Black = light_blue,
      White = light_red,
      Hispanic = light_green,
      Asian = strong_orange
    )
  ) +
  ggtitle(
    "Predominant Race and Ethnicity", 
    "2017 ACS 5-year estimates. Opacity = Percent."
  )

sts_2008 <- df_major %>% 
  filter(
    election == "primary", party == "DEMOCRATIC", 
    year == 2008, office == "SENATOR IN THE GENERAL ASSEMBLY",
    district == 1
  ) %>%
  group_by(ward, candidate) %>%
  summarise(votes = sum(votes)) %>%
  group_by(ward) %>%
  mutate(
    total_votes = sum(votes), 
    pvote=100 * votes/total_votes,
    candidate=format_name(candidate),
    is_winner = rank(desc(pvote)) == 1,
    margin = pvote - pvote[rank(desc(pvote)) == 2]
  )

district_map +
  geom_sf(
    data = st_intersection(wards, sts_1) %>% 
      inner_join(sts_2008 %>% filter(is_winner)),
    aes(fill=candidate, alpha=margin),
    color=NA,
  ) +
  scale_fill_manual(
    "Winner",
    values=c(light_green, light_purple)
  ) +
  scale_alpha_continuous("Margin (%)") +
  theme(legend.box = "vertical") +
  ggtitle("The 2008 PA-1 Democratic Primary")

races <- tribble(
  ~year, ~election, ~office, ~short_office, ~candidates,
  "2016", "primary", "PRESIDENT OF THE UNITED STATES", "President", list("HILLARY CLINTON", "BERNIE SANDERS"),
  "2017", "primary", "DISTRICT ATTORNEY", "District Attorney",list("LAWRENCE S KRASNER", "JOE KHAN"),
  "2019", "general", "COUNCIL AT LARGE", "City Council", list("KENDRA BROOKS", "DAVID OH")
)

div_cats <- readRDS("../../data/processed_data/div_cats_2019-12-03.RDS")
divs %<>% left_join(div_cats)

df_cand <- df_major %>% 
  inner_join(races) %>%
  filter(election == "general" | party == "DEMOCRATIC") %>%
  left_join(divs %>% as.data.frame() %>% select(warddiv, sts, cat)) %>%
  group_by(warddiv, year, office) %>%
  mutate(total_votes = sum(votes), pvote=100 * votes/total_votes) %>%
  filter(mapply(`%in%`, candidate, candidates))

df0 <- df_cand %>%
  group_by(candidate, year, election, short_office) %>%
  summarise(
    pvote_city = weighted.mean(pvote, total_votes),
    pvote_sts01 = weighted.mean(pvote, ifelse(sts == "001", total_votes, 0))
  ) %>%
  ungroup()

df0 %>%
  mutate(candidate = format_name(candidate)) %>% 
  arrange(year) %>%
  mutate(election = paste(year, format_name(election))) %>%
  select(election, candidate, pvote_city, pvote_sts01) %>%
  knitr::kable(digits=1, col.names=c("Election", "Candidate", "City %", "PA-1 %"))

bloc_colors <- c(
  "Black Voters" = light_blue,
  "Wealthy Progressives" = light_red,
  "White Moderates" = light_orange,
  "Hispanic North Philly" = light_green
)
district_map +
  geom_sf(
    data=divs,
    aes(fill = cat),
    color=NA,
    alpha=0.8
  ) +
  geom_sf(
    data=st_crop(sts, bb_phila), 
    fill=NA, 
    color="white"
  ) +
  scale_fill_manual(
    "Voting Bloc",
    values=bloc_colors
  ) +
  ggtitle("Voting Blocs of PA-1")

turnout_nums <- turnout %>% 
  filter(year %in% c(2016, 2017, 2019)) %>% 
  ungroup() %>% 
  inner_join(
    divs %>%
      filter(sts == "001") 
  ) %>% 
  group_by(year,election,cat) %>%
  summarise(turnout = sum(turnout)) %>% 
  group_by(year, election) %>%
  mutate(pturnout = turnout / sum(turnout)) %>% 
  as.data.frame()

cand_nums <- df_cand %>%
  group_by(year, election, short_office) %>%
  mutate(candnum = as.numeric(factor(candidate))) %>%
  select(year, election, short_office, candidate, candnum) %>%
  unique()

ggplot(
  df_cand %>% 
    ungroup() %>%
    filter(sts == "001") %>%
    group_by(candidate, year, election, short_office, cat) %>%
    summarise(votes = sum(votes)) %>%
    left_join(cand_nums)
) +
  geom_bar(
    aes(x = cat, y = votes, fill=cat),
    stat="identity"
  ) +
  scale_fill_manual(NULL, values=bloc_colors) +
  facet_grid(paste(year, format_name(election)) ~ candnum) +
  geom_label(
    data=cand_nums, 
    aes(label=format_name(candidate)), 
    x = 4.5, y=20e3,
    hjust=1,
    vjust=0
  ) +
  scale_y_continuous("Votes", labels = scales::comma) +
  theme_sixtysix() %+replace% 
  theme(
    strip.text.x = element_blank(),
    axis.text.x = element_blank()
  ) +
  labs(
    title="Recent results in PA-1",
    x=NULL
  )

district_map +
  geom_sf(
    data=divs %>% 
      left_join(df_cand) %>%
      # left_join(office_max) %>%
      left_join(cand_nums) %>%
      group_by(year, election, short_office, warddiv) %>%
      mutate(pvote_head2head = 100 * pvote / sum(pvote)),
    aes(fill=pvote_head2head),
    color=NA,
    alpha=0.8
  ) +
  geom_sf(
    data=st_crop(sts, bb_phila), 
    fill=NA, 
    color="white"
  ) +
  geom_label(
    data = cand_nums,
    aes(label=format_name(candidate)),
    hjust=1.1,
    vjust=1,
    x = bb["right"],
    y = bb["top"]
  ) +
  facet_grid(year + election ~ candnum) +
  theme(
    strip.text.x = element_blank(),
    legend.position = "right"
  ) +
  scale_fill_viridis_c("Head-to-head\n % of Vote") +
  ggtitle("Recent Elections in PA-1", "Head-to-head Vote Percentages")

prop_wealthy_progressive <- 0.7
line_slope <- -prop_wealthy_progressive / (1-prop_wealthy_progressive)
y_intercept <- 0.5 / (1-prop_wealthy_progressive)

cand_scatter <- df_cand %>% 
  ungroup() %>%
  filter(sts == "001") %>%
  mutate(
    is_wealthy = ifelse(
      cat == "Wealthy Progressives", "wealthyprogs", "other"
    )
  ) %>%
  group_by(candidate, year, election, short_office, is_wealthy) %>%
  summarise(votes = sum(votes)) %>%
  group_by(year, election, short_office, is_wealthy) %>%
  mutate(pvote_h2h = votes / sum(votes)) %>%
  ungroup() %>%
  select(year, election, short_office, candidate, is_wealthy, pvote_h2h) %>%
  spread(key=is_wealthy, value=pvote_h2h)
    

ggplot(
  cand_scatter,
  aes(x = 100*wealthyprogs, y=100*other)
) +
  geom_label(aes(label = format_name(candidate))) +
  geom_abline(
    slope = line_slope, 
    intercept = 100 * y_intercept,
    linetype="dashed"
  ) +
  coord_fixed() + 
  xlim(c(0,100)) +
  ylim(c(0, 100)) +
  annotate(
    "text",
    y=0,
    # x=-100*y_intercept/line_slope,
    x=0,
    label=sprintf("Assuming Wealthy Progressives have %s%% of turnout.\nCandidates are at their head-to-head percentages.", round(100*prop_wealthy_progressive)),
    hjust=0,
    vjust=-0.1,
    color="grey50"
  ) +
  labs(
    title="Win Line for PA-1",
    subtitle="Candidates to the top-right win.",
    x="Percent in Wealthy Progressive divisions",
    y="Percent in other divisions"
  ) +
  theme_sixtysix()