###These are the necessary functions:

###Libraries
#require (foreign)
require (ggplot2)

### 1.Function to make and format numbers.  It takes a character vector or a factor and it 
##  outputs a number. The digits argument specifies the number of digits.
make.number<-function (x, y) {x<-round(as.numeric(as.character(x)), digits=y)
                              return(x)}
### 2. Function to compute the mean government position for any variable 
##  The function takes as input the name of the Manifesto variable (party position).
##  The list is available here https://manifesto-project.wzb.eu/data/MPDataset_full_codebook.pdf
##  By default, the function computes the weighted average position (each party position is weighted by the share of the party from the total of the government coalition)
##  If weighted is switched off, a simple average of the governing parties' positions is computed 

party.position<-function (x, weighted=TRUE) {  
  t<-unique(cabinets$cabinet_id)         #get the list for all cabinets
  m<-matrix (NA, length(t), 3+length(x)) #construct a matrix that has as many columns as unique cabinets and four columns: three indicators for coutnry, start and end date, and the position itslef 
  for (i in 1:length(t)){                #for all cabinets
    j<-t[i]                              #get the cabinet id for ease
    temp<-subset(cabinets, cabinets$cabinet_id==j & cabinets$cabinet_party==1) #filter only the records relevant for this cabinet and the parties in the cabinet
    temp2<-subset(cabinets, cabinets$previous_cabinet_id==j)                   #filter only the records for the previous cabinet
    m[i,1]<-temp[1,colnames(temp)=='countryname']			         					   #country name
    m[i,2]<-temp[1,colnames(temp)=='start_date']        								   #start date
    m[i,3]<-temp2[1,colnames(temp)=='start_date']	        							   #end date: search for the record that has the current cabinet number as the previous cabinet, and put its starting date
    
    if (weighted==FALSE) {
      m[i,4]<-mean(temp[,colnames(temp)==x], na.rm=T)}				             #unweighted mean party position
    else {
      m[i,4]<-weighted.mean(temp[,colnames(temp)==x], temp$seats_share)}	 #weighted mean party position
  }
  
  m<-as.data.frame(m)
  colnames(m)<-c('Country','start.date','end.date',x)
  m<-subset(m,  is.na(m$Country)==F)
  m[,4]<-round(as.numeric(as.character((m[,4]))), digits=4)
  
  ###Format the dates of the data
  m$startY<-format(as.Date(m$start.date, format='%Y-%m-%d'),  format='%Y')      #cabinet start year
  m$startM<-format(as.Date(m$start.date, format='%Y-%m-%d'),  format='%m')      #cabinet start month
  m$startF<-format(as.Date(m$start.date, format='%Y-%m-%d'),  format='%Y.%m')   #cabinet start year/month
  
  m$endY<-format(as.Date(m$end.date, format='%Y-%m-%d'),  format='%Y')        	#cabinet end year
  m$endM<-format(as.Date(m$end.date, format='%Y-%m-%d'),  format='%m')	      	#cabinet end month
  m$endF<-format(as.Date(m$end.date, format='%Y-%m-%d'),  format='%Y.%m')	    	#cabinte end year/month
  
  
  return(m)
}


### 3. Function to prepare  a  list of year with the associated government positions
##  The function takes as input the data frame returned from the function party.position().
##  If  long=True a long data frame (county/year as the row) is returned. If long=F a wide data frame is returned with the coutnries as rows and the years  as  columns 
##  the start and end  argumetns specify the start and end years of the extraction  

party.position.yearly <-function (m, long=TRUE, start=1945, end=2010) {
  
  u.c<-unique(m$Country)[-c(5,17,27,32,33)]             #exclude non-Euroepan countries
  v<-seq(start,end, 1)                                  #start and end years
  year.table<-matrix(NA, length(u.c),length(v)+1 )      #prepare the matrix with rows the number of countries and columns the nubmer of years plus 1
  
  for (i in 1:length(u.c)) {                            #for each country    
    m.temp<-subset(m, m$Country==u.c[i])                #subset only the relevant observations
    year.table[i,1]<-as.character(u.c[i])               #record the name of the country
    for ( j in 1:length(v)) {                           #for each year  
      m.temp2<-subset(m.temp, m.temp$startY<=v[j] & m.temp$endY>=v[j] ) #subset only the cabinets that have started before and have no yet ended
      if (nrow(m.temp2)==0) next                        #if not applicabel move one
      else  
        if (nrow(m.temp2)==1)                           #if only one such cabinet
          year.table[i, j+1]<-m.temp2[1,4]              #get its position and record it
      else 
        if  (nrow(m.temp2)==2)
        {
          m.temp2<-m.temp2[order(m.temp2$endF) , ]
          z.v<-vector(mode='numeric',2)
          z.v[2]<-12-make.number(m.temp2$startM[2])
          z.v[1]<-12-z.v[2]
          year.table[i, j+1]<-sum(m.temp2[,4]*(z.v/12))}
      
      else    
        if (nrow(m.temp2)>1){                           #if more than 1 cabinet was active during the year
          m.temp2<-m.temp2[order(m.temp2$endF) , ]
          z<-nrow(m.temp2)                              #get their number
          z.v<-vector(mode='numeric',z)                 #make a vector of that length
          z.v[z]<-12-make.number(m.temp2$startM[z])     #record the number of months of the cabinet being active 
          for (k  in (z-1):2) {z.v[k]<-make.number(m.temp2$startM[k+1])-make.number(m.temp2$startM[k])}
          z.v[1]<-12-sum(z.v[-1])
          year.table[i, j+1]<-sum(m.temp2[,4]*z.v/12)}    #create the weighted yearly position
    }
  }
  year.table<-as.data.frame(year.table)
  colnames(year.table)<-c('Country',v)
  
  for (g in 2:NCOL(year.table)){year.table[,g]<-make.number(year.table[,g],2)}
  if (long){
    cab.table<-melt(year.table, id='Country')
    colnames(cab.table)<-c("State","Year","position")
    
    cab.table$id<-paste(cab.table$Year, cab.table$State, sep=".")}
  
  else
  {cab.table<-year.table}
  return (cab.table)
}


#### 4.The  same as above but on a monthly basis
party.position.monthly <-function (m, long=TRUE, start=1945, end=2010) {
  
  u.c<-unique(m$Country)[-c(5,17,27,32,33)]             #exclude non-Euroepan countries
  v1<-seq(start,end, 1)                                  #start and end years
  v2<-seq(0.01,0.12,0.01)
  v<-rep(NA, length(v1)*12)
  for (i in 1:length(v1)) {v[(1+12*(i-1)):(12+12*(i-1))]<-v1[i]+v2}
  
  year.table<-matrix(NA, length(u.c),length(v)+1 )      #prepare the matrix with rows the number of countries and columns the nubmer of years plus 1
  
  for (i in 1:length(u.c)) {                            #for each country    
    m.temp<-subset(m, m$Country==u.c[i])                #subset only the relevant observations
    year.table[i,1]<-as.character(u.c[i])               #record the name of the country
    for ( j in 1:length(v)) {                           #for each year  
      m.temp2<-subset(m.temp, m.temp$startF<=v[j] & m.temp$endF>v[j] ) #subset only the cabinets that have started before and have no yet ended
      if (nrow(m.temp2)==0) next                        #if not applicable move on
      else  
        
        year.table[i, j+1]<-m.temp2[1,4]              #get its position and record it
    }
  }
  year.table<-as.data.frame(year.table)
  colnames(year.table)<-c('Country',v)
  
  for (g in 2:NCOL(year.table)){year.table[,g]<-make.number(year.table[,g],2)}
  if (long){
    cab.table<-melt(year.table, id='Country')
    colnames(cab.table)<-c("State","Year.month","position")
    
    cab.table$id<-paste(cab.table$Year.month, cab.table$State, sep=".")}
  
  else
  {cab.table<-year.table}
  return (cab.table)
}



#### 5.Master function for the extraction of manifesto positions for governments.
## Requires the functions party.position() and party.position.yearly()
##  The function takes as input the name of the Manifesto variable (party position).
##  The list is available here https://manifesto-project.wzb.eu/data/MPDataset_full_codebook.pdf
##  By default, the function computes the weighted average position (each party position is weighted by the share of the party from the total of the government coalition)
##  If weighted is switched off, a simple average of the governing parties' positions is computed 
##  If  long=True a long data frame (county/year as the row) is returned. If long=F a wide data frame is returned with the coutnries as rows and the years  as  columns 
##  the start and end  argumetns specify the start and end years of the extraction  
##  the argument 'period' has two values - 'year' (default) and 'month' which extracts the positions for every month
##  N.B. the file 'end dates for the cabinets2.r' is needed to supply the end date of the last governments featured in the Manifesto database 

manifesto.position <-function (x, weighted=T, period='year', long=T, start=1945, end=2010) { 
  
  m<-party.position (x, weighted=weighted)
  assign("m", m, .GlobalEnv)
  #source('end dates for the cabinets2.r')
  source ('http://www.dimiter.eu/Data_files/cabinets/end%20dates%20for%20the%20cabinets2.r')
  assign("m", m, .GlobalEnv)
  if(period=='month')  {
    t<-party.position.monthly(m, long=long, start=start, end=end) }
  else  
    t<-party.position.yearly(m, long=long, start=start, end=end)
  
  return(t)
  
  
}