library(xtable)
library(dplyr)
library(tidyverse)
if (interactive()){
  print("Interactive Session, using home directories")
  data_dir = "~/git/"
} else {
  data_dir = "/srv/shiny-server/git/"
}

latex_outdir = "~/git/Writing/thesis/tex/appendices/"
thing  = read.csv(paste0(data_dir, "Data/ml_combined/genes/pnca_70_30_actual.csv"))

# list of splits
split_type = c(
  # "cd_7030", 
  # "cd_8020", 
  "cd_sl"
  # "none"
)
split_choicenames=c(
  # "70:30", 
  # "80:20", 
  "Scaling law"
  # "CV thresholds"
)
split_file = c(
  # "_70_30_complete", 
  # "_80_20_complete", 
  "_sl_complete"
  # "_none_complete"
)

split_file_FS = c(
  # "_70_30_complete", 
  # "_80_20_complete", 
  "_sl_complete"
)

# necessary because the names will be wrong otherwise
split_map = data.frame(
  files=c(
    # "_70_30_complete",
    # "_80_20_complete",
    "_sl_complete"
    # "_none_complete"
  ),
  splits=c(
    # "cd_7030",
    # "cd_8020",
    "cd_sl"
    # "none"
  )
)

metadata_cols = c("n_training_size", "n_test_size", "n_trainingY_ratio", "n_testY_ratio", "resampling", "n_features")

# hardcoded list of drugs
drug = c("ethambutol", "isoniazid", "pyrazinamide", "rifampicin", "streptomycin")

drug_choicenames = c("EmbB-ethambutol", "KatG-isoniazid", "PncA-pyrazinamide", "RpoB-rifampicin", "GidB-streptomycin")

gene = c("embb", "katg", "pnca", "rpob", "gid")
combo = data.frame(drug, gene)

score_cols = c(
  'MCC',
  'F1',
  'Accuracy',
  'JCC',
  'Recall',
  'Precision',
  'source_data',
  'Model_name'
)


# Loader for per-gene CSVs
#"/home/sethp/git/Data/ml_combined/genes/pnca_70_30_complete.csv"
loaded_files=list()
for (x in gene) {
  #x=tolower(x)
  for (split in split_file){
    filedata = paste0(x, split)
    filename = paste0(data_dir,'LSHTM_ML/output/genes/',x,split,'.csv')
    
    #print(c(filename))
    #load_name=paste0(combo[gene==x,"drug"],'_',split_map['splits'][split_map['files']==split])
    load_name=paste0(x,'_baselineC_',split_map['splits'][split_map['files']==split])
    print(load_name)
    # try() on its own is fine here because we don't need to do anything if it fails
    try({loaded_files[[load_name]] = read.csv(filename)})
  }
}
# Loader for per-gene Feature Selection CSVs
for (x in gene) {
  #x=tolower(x)
  for (split in split_file_FS){
    filedata = paste0(x, split)
    filename = paste0(data_dir,'LSHTM_ML/output/genes/',x,split,'_FS.csv')
    
    #print(c(filename))
    #load_name=paste0(combo[gene==x,"drug"],'_',split_map['splits'][split_map['files']==split])
    load_name=paste0(x,'_baselineC_',split_map['splits'][split_map['files']==split], '_FS')
    print(load_name)
    # try() on its own is fine here because we don't need to do anything if it fails
    try({loaded_files[[load_name]] = read.csv(filename)})
  }
}


# Funky loader for combined data
for (x in gene) {
  for (ac in c('_complete', '_FS')){
    for (gene_count in c(1:5)){
      load_name=paste0(gene_count, "genes_logo_skf_BT_", x, ac)
      filename = paste0(data_dir,'LSHTM_ML/output/combined/',load_name, ".csv")
      store_name=paste0(gene_count, "genes_logo_skf_BT_", x, ac)
      # tryCatch is necessary here rather than try() because we need to do more
      # manipulation afterwards (throwing away the column after loading)
      load_successful=TRUE
      tryCatch({temp_df = read.csv(filename)},error=function(e){load_successful<<-FALSE})
      if (load_successful){
        temp_df=temp_df[, 2:ncol(temp_df)] # throw away first column
        loaded_files[[store_name]] = temp_df
        print(paste0("loaded file: ", filename, "into var: ", store_name))
      }
    }
  }
}

loaded_files$embb_baselineC_cd_sl = loaded_files$embb_baselineC_cd_sl[loaded_files$embb_baselineC_cd_sl$resampling=='none',score_cols]
loaded_files$katg_baselineC_cd_sl = loaded_files$katg_baselineC_cd_sl[loaded_files$katg_baselineC_cd_sl$resampling=='none',score_cols]
loaded_files$pnca_baselineC_cd_sl = loaded_files$pnca_baselineC_cd_sl[loaded_files$pnca_baselineC_cd_sl$resampling=='none',score_cols]
loaded_files$rpob_baselineC_cd_sl = loaded_files$rpob_baselineC_cd_sl[loaded_files$rpob_baselineC_cd_sl$resampling=='none',score_cols]
loaded_files$gid_baselineC_cd_sl  = loaded_files$gid_baselineC_cd_sl[loaded_files$gid_baselineC_cd_sl$resampling=='none',score_cols]

loaded_files$embb_baselineC_cd_sl_FS = loaded_files$embb_baselineC_cd_sl_FS[loaded_files$embb_baselineC_cd_sl_FS$resampling=='none',score_cols]
loaded_files$katg_baselineC_cd_sl_FS = loaded_files$katg_baselineC_cd_sl_FS[loaded_files$katg_baselineC_cd_sl_FS$resampling=='none',score_cols]
loaded_files$pnca_baselineC_cd_sl_FS = loaded_files$pnca_baselineC_cd_sl_FS[loaded_files$pnca_baselineC_cd_sl_FS$resampling=='none',score_cols]
loaded_files$rpob_baselineC_cd_sl_FS = loaded_files$rpob_baselineC_cd_sl_FS[loaded_files$rpob_baselineC_cd_sl_FS$resampling=='none',score_cols]
loaded_files$gid_baselineC_cd_sl_FS = loaded_files$gid_baselineC_cd_sl_FS[loaded_files$gid_baselineC_cd_sl_FS$resampling=='none',score_cols]

loaded_files$`5genes_logo_skf_BT_embb_complete` = loaded_files$`5genes_logo_skf_BT_embb_complete`[loaded_files$`5genes_logo_skf_BT_embb_complete`$resampling=='none',score_cols]
loaded_files$`5genes_logo_skf_BT_katg_complete` = loaded_files$`5genes_logo_skf_BT_katg_complete`[loaded_files$`5genes_logo_skf_BT_katg_complete`$resampling=='none',score_cols]
loaded_files$`5genes_logo_skf_BT_pnca_complete` = loaded_files$`5genes_logo_skf_BT_pnca_complete`[loaded_files$`5genes_logo_skf_BT_pnca_complete`$resampling=='none',score_cols]
loaded_files$`5genes_logo_skf_BT_rpob_complete` = loaded_files$`5genes_logo_skf_BT_rpob_complete`[loaded_files$`5genes_logo_skf_BT_rpob_complete`$resampling=='none',score_cols]
loaded_files$`5genes_logo_skf_BT_gid_complete` = loaded_files$`5genes_logo_skf_BT_gid_complete`[loaded_files$`5genes_logo_skf_BT_gid_complete`$resampling=='none',score_cols]

loaded_files$`5genes_logo_skf_BT_embb_complete`[loaded_files$`5genes_logo_skf_BT_embb_complete`$source_data=="Train","source_data"]="0Train"
loaded_files$`5genes_logo_skf_BT_katg_complete`[loaded_files$`5genes_logo_skf_BT_katg_complete`$source_data=="Train","source_data"]="0Train"
loaded_files$`5genes_logo_skf_BT_pnca_complete`[loaded_files$`5genes_logo_skf_BT_pnca_complete`$source_data=="Train","source_data"]="0Train"
loaded_files$`5genes_logo_skf_BT_rpob_complete`[loaded_files$`5genes_logo_skf_BT_rpob_complete`$source_data=="Train","source_data"]="0Train"
loaded_files$`5genes_logo_skf_BT_gid_complete`[loaded_files$`5genes_logo_skf_BT_gid_complete`$source_data=="Train","source_data"]="0Train"

loaded_files$`1genes_logo_skf_BT_embb_FS` = loaded_files$`1genes_logo_skf_BT_embb_FS`[loaded_files$`1genes_logo_skf_BT_embb_FS`$resampling=='none',score_cols]
loaded_files$`1genes_logo_skf_BT_katg_FS` = loaded_files$`1genes_logo_skf_BT_katg_FS`[loaded_files$`1genes_logo_skf_BT_katg_FS`$resampling=='none',score_cols]
loaded_files$`1genes_logo_skf_BT_pnca_FS` = loaded_files$`1genes_logo_skf_BT_pnca_FS`[loaded_files$`1genes_logo_skf_BT_pnca_FS`$resampling=='none',score_cols]
loaded_files$`1genes_logo_skf_BT_rpob_FS` = loaded_files$`1genes_logo_skf_BT_rpob_FS`[loaded_files$`1genes_logo_skf_BT_rpob_FS`$resampling=='none',score_cols]
loaded_files$`1genes_logo_skf_BT_gid_FS` = loaded_files$`1genes_logo_skf_BT_gid_FS`[loaded_files$`1genes_logo_skf_BT_gid_FS`$resampling=='none',score_cols]

loaded_files$`1genes_logo_skf_BT_embb_FS`[loaded_files$`1genes_logo_skf_BT_embb_FS`$source_data=="Train","source_data"]="0Train"
loaded_files$`1genes_logo_skf_BT_katg_FS`[loaded_files$`1genes_logo_skf_BT_katg_FS`$source_data=="Train","source_data"]="0Train"
loaded_files$`1genes_logo_skf_BT_pnca_FS`[loaded_files$`1genes_logo_skf_BT_pnca_FS`$source_data=="Train","source_data"]="0Train"
loaded_files$`1genes_logo_skf_BT_rpob_FS`[loaded_files$`1genes_logo_skf_BT_rpob_FS`$source_data=="Train","source_data"]="0Train"
loaded_files$`1genes_logo_skf_BT_gid_FS`[loaded_files$`1genes_logo_skf_BT_gid_FS`$source_data=="Train","source_data"]="0Train"

# Rewrite CV/BT to Train/Test and ensure that Train gets presented first
loaded_files$embb_baselineC_cd_sl_FS[loaded_files$embb_baselineC_cd_sl_FS$source_data=="CV","source_data"]="0Train"
loaded_files$katg_baselineC_cd_sl_FS[loaded_files$katg_baselineC_cd_sl_FS$source_data=="CV","source_data"]="0Train"
loaded_files$pnca_baselineC_cd_sl_FS[loaded_files$pnca_baselineC_cd_sl_FS$source_data=="CV","source_data"]="0Train"
loaded_files$rpob_baselineC_cd_sl_FS[loaded_files$rpob_baselineC_cd_sl_FS$source_data=="CV","source_data"]="0Train"
loaded_files$gid_baselineC_cd_sl_FS[loaded_files$gid_baselineC_cd_sl_FS$source_data=="CV","source_data"]="0Train"
loaded_files$embb_baselineC_cd_sl_FS[loaded_files$embb_baselineC_cd_sl_FS$source_data=="BT","source_data"]="Test"
loaded_files$katg_baselineC_cd_sl_FS[loaded_files$katg_baselineC_cd_sl_FS$source_data=="BT","source_data"]="Test"
loaded_files$pnca_baselineC_cd_sl_FS[loaded_files$pnca_baselineC_cd_sl_FS$source_data=="BT","source_data"]="Test"
loaded_files$rpob_baselineC_cd_sl_FS[loaded_files$rpob_baselineC_cd_sl_FS$source_data=="BT","source_data"]="Test"
loaded_files$gid_baselineC_cd_sl_FS[loaded_files$gid_baselineC_cd_sl_FS$source_data=="BT","source_data"]="Test"

loaded_files$embb_baselineC_cd_sl[loaded_files$embb_baselineC_cd_sl$source_data=="Train","source_data"]="0Train"
loaded_files$katg_baselineC_cd_sl[loaded_files$katg_baselineC_cd_sl$source_data=="Train","source_data"]="0Train"
loaded_files$pnca_baselineC_cd_sl[loaded_files$pnca_baselineC_cd_sl$source_data=="Train","source_data"]="0Train"
loaded_files$rpob_baselineC_cd_sl[loaded_files$rpob_baselineC_cd_sl$source_data=="Train","source_data"]="0Train"
loaded_files$gid_baselineC_cd_sl[loaded_files$gid_baselineC_cd_sl$source_data=="Train","source_data"]="0Train"

# Normal tables
for (x in gene) {
  out_tex=paste0(latex_outdir,"ml_table-", x, ".tex")
  df_name=paste0(x,"_baselineC_cd_sl")
  processed_df = loaded_files[[df_name]] %>% pivotf_wider(names_from=source_data, values_from=c(
    'MCC',
    'F1',
    'Accuracy',
    'Recall',
    'Precision',
    'JCC'
  ),
  names_sep = " ",
  names_sort = TRUE
  )
  processed_df=as.data.frame(processed_df)
  
  colnames(processed_df)=gsub('0Train','Train',colnames(processed_df))
  rownames(processed_df)=processed_df$Model_name
  #colnames(processed_df)=gsub('Model_name','Model name',colnames(processed_df))
  processed_df=subset(processed_df, select=-Model_name)
  assign(paste0(x,"_out"),processed_df)
  print(out_tex)
  print.xtable(xtable(
    processed_df, type="latex"),
    file=out_tex
  )
}

quick_latex(huxtable(processed_df), file="/tmp/foo.tex", open=FALSE)


# Feature Selection tables
for (x in gene) {
  out_tex=paste0(latex_outdir,"ml_table_FS-", x, ".tex")
  df_name=paste0(x,"_baselineC_cd_sl_FS")
  processed_df = loaded_files[[df_name]] %>% pivot_wider(names_from=source_data, values_from=c(
    'MCC',
    'F1',
    'Accuracy',
    'Recall',
    'Precision',
    'JCC'
  ),
  names_sep = " ",
  names_sort = TRUE
  )
  processed_df=as.data.frame(processed_df)
  
  colnames(processed_df)=gsub('0Train','Train',colnames(processed_df))
  rownames(processed_df)=processed_df$Model_name
  #colnames(processed_df)=gsub('Model_name','Model name',colnames(processed_df))
  processed_df=subset(processed_df, select=-Model_name)
  assign(paste0(x,"_out"),processed_df)
  print(out_tex)
  print.xtable(xtable(
    processed_df, type="latex"),
    file=out_tex
  )
}

# Combined Normal
for (x in gene) {
  out_tex=paste0(latex_outdir,"ml_table_combined-", x, ".tex")
  df_name=paste0("5genes_logo_skf_BT_",x,"_complete")
  processed_df = loaded_files[[df_name]] %>% pivot_wider(names_from=source_data, values_from=c(
    'MCC',
    'F1',
    'Accuracy',
    'Recall',
    'Precision',
    'JCC'
  ),
  names_sep = " ",
  names_sort = TRUE
  )
  processed_df=as.data.frame(processed_df)
  
  colnames(processed_df)=gsub('0Train','Train',colnames(processed_df))
  rownames(processed_df)=processed_df$Model_name
  #colnames(processed_df)=gsub('Model_name','Model name',colnames(processed_df))
  processed_df=subset(processed_df, select=-Model_name)
  assign(paste0(x,"_out"),processed_df)
  print.xtable(xtable(
    processed_df, type="latex"),
    file=out_tex
  )
}

# Combined Feature Selection
for (x in gene) {
  out_tex=paste0(latex_outdir,"ml_table_combined-FS-", x, ".tex")
  df_name=paste0("1genes_logo_skf_BT_",x,"_FS")
  processed_df = loaded_files[[df_name]] %>% pivot_wider(names_from=source_data, values_from=c(
    'MCC',
    'F1',
    'Accuracy',
    'Recall',
    'Precision',
    'JCC'
  ),
  names_sep = " ",
  names_sort = TRUE
  )
  processed_df=as.data.frame(processed_df)
  
  colnames(processed_df)=gsub('0Train','Train',colnames(processed_df))
  rownames(processed_df)=processed_df$Model_name
  #colnames(processed_df)=gsub('Model_name','Model name',colnames(processed_df))
  processed_df=subset(processed_df, select=-Model_name)
  assign(paste0(x,"_out"),processed_df)
  print(out_tex)
  print.xtable(xtable(
    processed_df, type="latex"),
    file=out_tex
  )
}