/Users/barrand/private/dev/softinex/old/inexlib-1.2/inlib/inlib/rcsv_ntuple

Go to the documentation of this file.

// Copyright (C) 2010, Guy Barrand. All rights reserved.
// See the file inlib.license for terms.

#ifndef inlib_rcsv_ntuple
#define inlib_rcsv_ntuple

// A simple ntuple class to read at the csv format.
// (csv = comma separated value).

// This reader can be use to read file at the hippodraw format
// which is :
// - one header line for the ntuple title.
// - one csv line for column names.
// - data at csv format.

#include "rntuple"

#include <istream>
#include <sstream>

#include "vfind"
#include "vmanip"
#include "words"
#include "sto"
#include "s2time"
#include "chars"
#include "strip"
#include "cids"

#ifdef INLIB_MEM
#include "mem"
#endif

namespace inlib {
namespace rcsv {

class ntuple : public virtual read::intuple {
public: //read::intuple
  virtual void start() {
    m_reader.clear();
    m_reader.seekg(0,std::ios::beg);
    if(m_hippo) {
      skip_line(m_reader,m_sz);
      skip_line(m_reader,m_sz);
    }
  }
  virtual bool next() { 
    if(!m_sep) return false; //not inited.
    if(m_reader.tellg()>=m_sz) return false;
    // first time we are at bol but else we are at eol.
    char c;
    m_reader.get(c);
    if(c==LF()){
      if(m_reader.tellg()>=m_sz) {
        //eof. Tell caller to stop looping on ntuple rows.
        return false;
      }
      //eol. Next char read is going to be at bol.
    } else {
      m_reader.putback(c);
      //bol
    }
    // ready for a new row :

    while(skip_comment(m_reader,m_sz)){}
    if(m_reader.tellg()>=m_sz) return false;

    return _read_line();
  }

  virtual read::icol* find_icol(const std::string& a_name){
    return find_named<read::icol>(m_cols,a_name);
  }

  virtual const std::vector<read::icol*>& columns() const {return m_cols;}
protected:
  static bool _read(std::istream& a_reader,
                           char,std::streampos,
                           double& a_v) {
    a_reader >> a_v;
    if(a_reader.tellg()==std::streampos(-1)) {a_v = 0;return false;}
    //std::cout << "debug : _read(double) " << a_v << std::endl;
    return true;
  }
  static bool _read(std::istream& a_reader,
                           char a_sep,std::streampos a_sz,
                           time_t& a_v) {
    std::string s;
    char c;
    while(true){
      if(a_reader.tellg()>=a_sz) break;
      a_reader.get(c);
      if((c==a_sep)||(c==CR())||(c==LF())) {
        a_reader.putback(c);
        break;
      }
      s += c;
    }
    if(!s2time(s,a_v)) return false;
    return true;
  }
  static bool _read(std::istream& a_reader,
                           char a_sep,std::streampos a_sz,
                           std::string& a_v) {
    a_v.clear();
    char c;
    while(true){
      if(a_reader.tellg()>=a_sz) break;
      a_reader.get(c);
      if((c==a_sep)||(c==CR())||(c==LF())) {
        a_reader.putback(c);
        break;
      }
      a_v += c;
    }
    return true;
  }

public:
  template <class T>
  class column : public virtual read::icolumn<T> {
  public: //icol
    virtual const std::string& name() const {return m_name;}
  public: //icolumn<T>
    virtual bool get_entry(T& a_v) const {
      a_v = m_tmp;
      return true;
    }
  public:
    column(const std::string& a_name)
    :m_name(a_name)
    ,m_tmp(T())
    {}
    virtual ~column(){}
  protected:
    column(const column& a_from)
    :read::intuple(a_from),read::icolumn<T>(a_from)
    ,m_name(a_from.m_name) 
    ,m_tmp(a_from.m_tmp)
    {}
    column& operator=(const column& a_from){
      m_name = a_from.m_name;
      m_tmp = a_from.m_tmp;
      return *this;
    }
  public:
    // should be used in ntuple _read_line only :
    void set_value(const T& a_v){m_tmp = a_v;}
  protected:
    std::string m_name;
    T m_tmp;
  };


#ifdef INLIB_MEM
public:
  static const std::string& s_class() {
    static const std::string s_v("inlib::rcsv::ntuple");
    return s_v;
  }
#endif
public:
  ntuple(std::istream& a_reader)
  :m_reader(a_reader)
  ,m_sep(0)
  ,m_sz(0)
  ,m_hippo(false)
  {
#ifdef INLIB_MEM
    mem::increment(s_class().c_str());
#endif
  }
  virtual ~ntuple() {
    inlib::clear<read::icol>(m_cols);
#ifdef INLIB_MEM
    mem::decrement(s_class().c_str());
#endif
  }
protected:
  ntuple(const ntuple& a_from)
  :read::intuple(a_from)
  ,m_reader(a_from.m_reader)
  ,m_sep(a_from.m_sep)
  ,m_sz(a_from.m_sz)
  ,m_hippo(a_from.m_hippo)
  {
#ifdef INLIB_MEM
    mem::increment(s_class().c_str());
#endif
  }
  ntuple& operator=(const ntuple& a_from){
    m_sep = a_from.m_sep;
    m_hippo = a_from.m_hippo;
    return *this;
  }
public:
  void set_hippo(bool a_hippo) {m_hippo = a_hippo;}

  std::istream& istrm() {return m_reader;}

  static bool find_sep(std::ostream& a_out,
                              std::istream& a_reader,bool a_hippo,
                              bool a_verbose,
                              char& a_sep){
    // analyse first data line to find the char separator.

    a_reader.clear();
    a_reader.seekg(0,std::ios::end);
    std::streampos sz = a_reader.tellg();
    a_reader.seekg(0,std::ios::beg);
    if(!sz) {
      a_out << "inlib::rcsv::ntuple::find_sep :"
            << " stream is empty."
            << std::endl;
      a_sep = 0;
      return false;
    } //file empty.
    if(a_verbose) a_out << "file size " << sz << std::endl;

    if(a_hippo) { //skip first two lines :
      if(!skip_line(a_reader,sz)) {a_sep = 0;return false;}
      if(!skip_line(a_reader,sz)) {a_sep = 0;return false;}
    } else {
      while(skip_comment(a_reader,sz)){}
    }
    if(a_reader.tellg()>=sz) {a_sep=0;return false;} //no data line.

    // get first data line :
    std::string sfirst;
   {char c;
    while(true) {
      if(a_reader.tellg()>=sz) break;
      a_reader.get(c);
      if((c==CR())||(c==LF())) break;
      sfirst += c;
    }}
    if(sfirst.empty()) {
      a_out << "inlib::rcsv::ntuple::find_set :"
            << " first datat line is empty."
            << std::endl;
      a_sep = 0;
      return false;
    }
    if(a_verbose) a_out << "first data line \"" << sfirst << "\"" << std::endl;

    //guess sep from first data line :
    std::istringstream strm(sfirst.c_str());
    double d;
    strm >> d;
    std::streampos pos = strm.tellg();
    if(pos==std::streampos(-1)) {
      a_out << "inlib::rcsv::ntuple::find_sep :"
            << " first line does not start with a number."
            << std::endl;
      a_sep = 0;
      return false;
    } //not a number.
    if(a_verbose) a_out << "first number " << d
                        << " ending at pos " << pos << std::endl;
    if(pos>=(std::streampos)sfirst.size()) {
      a_out << "inlib::rcsv::ntuple::find_sep :"
            << " no separator found in first line."
            << " pos " << pos
            << " sfirst.size() " << sfirst.size()
            << std::endl;
      a_sep = 0;
      return false;
    } //no sep.

    strm.get(a_sep);

    return true;
  }

public:
  bool initialize(std::ostream& a_out,
                         char a_sep = 0, //guessed
                         const std::string& a_suffix = "x", //col suffix
                         bool a_verbose = false) {
    inlib::clear<read::icol>(m_cols);
    m_sep = 0;
    m_sz = 0;

    if(a_suffix.empty()) {
      a_out << "inlib::rcsv::ntuple::initialize :"
            << " expect a column suffix."
            << std::endl;
      return false;
    }

    m_reader.clear();
    m_reader.seekg(0,std::ios::end);
    m_sz = m_reader.tellg();
    m_reader.seekg(0,std::ios::beg);
    if(!m_sz) {
      a_out << "inlib::rcsv::ntuple::initialize :"
            << " stream is empty."
            << std::endl;
      return false; //file empty.
    }
    if(a_verbose) a_out << "file size " << m_sz << std::endl;

    std::vector<std::string> labels;
    if(m_hippo) { //skip first two lines :
      std::string title;
      if(!read_line(m_reader,m_sz,title)) {a_sep = 0;return false;}
      std::string s;
      if(!read_line(m_reader,m_sz,s)) {a_sep = 0;return false;}
      inlib::words(s,"\t",false,labels);
    } else {
      while(skip_comment(m_reader,m_sz)){}
    }
    if(m_reader.tellg()>=m_sz) {m_sz=0;return false;}

    // get first data line :
    std::string sfirst;
  {{char c;
    while(true) {
      if(m_reader.tellg()>=m_sz) break;
      m_reader.get(c);
      if((c==CR())||(c==LF())) break;
      sfirst += c;
    }}
    if(sfirst.empty()) {
      a_out << "inlib::rcsv::ntuple::initialize :"
            << " first datat line is empty."
            << std::endl;
      m_sz = 0;
      return false;
    }}
    if(a_verbose) a_out << "first data line \"" << sfirst << "\"" << std::endl;

    if(a_sep) {
      m_sep = a_sep;
    } else {
      //guess sep from first data line :
      std::istringstream strm(sfirst.c_str());
      double d;
      strm >> d;
      std::streampos pos = strm.tellg();
      if(pos==std::streampos(-1)) {
        a_out << "inlib::rcsv::ntuple::initialize :"
              << " first line does not start with a number."
              << std::endl;
        m_sz = 0;
        return false;
      }
      if(a_verbose) a_out << "first number " << d
                          << " ending at pos " << pos << std::endl;
      if(pos>=(std::streampos)sfirst.size()) {
        a_out << "inlib::rcsv::ntuple::initialize :"
              << " no separator found in first line."
              << std::endl;
        m_sz = 0;
        return false;
      }
      strm.get(m_sep);
    }
    if(a_verbose) a_out << "sep " << (int)m_sep << std::endl;

    // in case sep is ' ', there is an ambiguity with some leading
    // space in front of first number.
    if(m_sep==' ') inlib::strip(sfirst,leading,' ');

    std::vector<std::string> words;
   {std::string sep;
    sep += m_sep;
    inlib::words(sfirst,sep,true,words);}

    // look if words are numbers :
    if(a_verbose) a_out << "words " << words.size() << std::endl;
    unsigned int index = 0;
    std::vector<std::string>::iterator it;
    for(it=words.begin();it!=words.end();++it,index++) {
      if(a_verbose) a_out << "word " << sout(*it) << "" << std::endl;
      if((*it).empty()) {
        // do not accept :
        //   <num><sep><num><sep><sep><num>...
        // but accept a trailing <sep> (glast.tnt) :
        //   <num><sep><num>....<sep><num><sep>
        if(index==(words.size()-1)) {
          break;
        } else {
          a_out << "inlib::rcsv::ntuple::initialize :"
                << " empty word."
                << std::endl;
          m_sep = 0;
          m_sz = 0;
          return false;
        }      
      }      
      std::string name(a_suffix+to<unsigned int>(m_cols.size()));
      if(m_hippo) {
        if(index>=labels.size()) {
          a_out << "inlib::rcsv::ntuple::initialize :"
                << " warning : not enough labels."
                << std::endl;
        } else {
          name = labels[index];
        }
      }
      double d;
      if(to<double>(*it,d)) {
        if(a_verbose) a_out << "number " << d << std::endl;
        create_column<double>(name);
      } else {
        time_t time;
        if(s2time(*it,time)) {
          create_column<time_t>(name);
        } else {
          create_column<std::string>(name);
        }
      }
    }
    unsigned int num = m_cols.size();
    if(!num) {
      a_out << "inlib::rcsv::ntuple::initialize :"
            << " zero columns."
            << std::endl;
      m_sep = 0;
      m_sz = 0;
      return false;
    }

    return true;
  }

  static const std::string& s_cid(cid a_id) {
    if(a_id==_cid(double())) {
      static const std::string s_v("double");
      return s_v;      
    } else if(a_id==_cid(time_t())) {
      static const std::string s_v("time");
      return s_v;      
    } else if(a_id==_cid(std::string())) {
      static const std::string s_v("string");
      return s_v;      
    } else {
      static const std::string s_v("unknown");
      return s_v;      
    }
  }

  void dump_columns(std::ostream& a_out) const {
    if((m_sep>=32)&&(m_sep<=126)) { //printable
      a_out << "separator is '" << m_sep << "'" << std::endl;
    } else {
      a_out << "separator is " << (unsigned int)m_sep << std::endl;
    }
    std::vector<read::icol*>::const_iterator it;
    for(it=m_cols.begin();it!=m_cols.end();++it) {
      a_out << (*it)->name()
            << " " << s_cid((*it)->id_cls())
            << std::endl;
    }
  }

protected:
  template <class T>
  column<T>* create_column(const std::string& a_name){
    if(find_named<read::icol>(m_cols,a_name)) return 0;
    column<T>* col = new column<T>(a_name);
    if(!col) return 0;
    m_cols.push_back(col);
    return col;
  }

  static bool read_line(std::istream& a_reader,std::streampos a_sz,
                               std::string& a_s){
    a_s.clear();
    char c;
    while(true) {
      if(a_reader.tellg()>=a_sz) {a_s.clear();return false;}
      a_reader.get(c);
      if(c==CR()) continue;
      if(c==LF()) break; //eol.
      a_s += c;
    }
    return true;
  }

  static bool skip_line(std::istream& a_reader,std::streampos a_sz){
    char c;
    while(true) {
      if(a_reader.tellg()>=a_sz) return false;
      a_reader.get(c);
      if(c==LF()) break;
    }
    return true;
  }

  static bool skip_comment(std::istream& a_reader,std::streampos a_sz){
    //ret true = we had a commented line, false : a data line or nothing.
    if(a_reader.tellg()>=a_sz) return false;
    //we should be at bol :
    char c;
    a_reader.get(c);
    if(c=='#') {
      return skip_line(a_reader,a_sz);
      //eol. Next char should be bol.
    } else {
      a_reader.putback(c);
      return false;
    }
  }

  bool _read_line() {
    // have to loop on all columns !
    typedef read::icol icol_t;
    typedef ntuple::column<double> cold_t;
    typedef ntuple::column<time_t> colt_t;
    typedef ntuple::column<std::string> cols_t;
    unsigned int index = 0;
    unsigned int num = m_cols.size();
    std::vector<icol_t*>::const_iterator it;
    for(it=m_cols.begin();it!=m_cols.end();++it,index++) {
      if(cold_t* cold = inlib::id_cast<icol_t,cold_t>(*(*it))) {
        double v;
        if(!_read(m_reader,m_sep,m_sz,v)) return false;
        cold->set_value(v);
      } else if(colt_t* colt = inlib::id_cast<icol_t,colt_t>(*(*it))) {
        time_t v;
        if(!_read(m_reader,m_sep,m_sz,v)) return false;
        colt->set_value(v);
      } else if(cols_t* cols = inlib::id_cast<icol_t,cols_t>(*(*it))) {
        std::string v;
        if(!_read(m_reader,m_sep,m_sz,v)) return false;
        cols->set_value(v);
      } else {
        //std::cout << "column cast failed." << std::endl;
        return false; 
      }
      if(index==(num-1)) { //read up to LF()
        char c;
        while(true){
          if(m_reader.tellg()>=m_sz) break;
          m_reader.get(c);
          if(c==LF()) break;
        }
      } else { //read sep :
        char sep;
        m_reader.get(sep);
      }
    }
    return true;
  }
protected:
  std::istream& m_reader;
  char m_sep;
  std::vector<read::icol*> m_cols;
  std::streampos m_sz;
  bool m_hippo;
};

}}


#include <fstream>

namespace inlib {
namespace rcsv {

class fntuple : public ntuple {
public:
  fntuple(const std::string& a_file)
  :ntuple(m_freader)
  ,m_file(a_file)
  {}
  virtual ~fntuple() {m_freader.close();}
protected:
  fntuple(const fntuple& a_from)
  :read::intuple(a_from),ntuple(a_from)
  ,m_file(a_from.m_file)
  {}
  fntuple& operator=(const fntuple& a_from){
    m_file = a_from.m_file;
    return *this;
  }
public:
  bool open(){
    m_freader.open(m_file.c_str());
    return m_freader.fail()?false:true;
  }
  bool initialize(std::ostream& a_out,
                         char a_sep = 0, //guessed
                         const std::string& a_suffix = "x", //col suffix
                         bool a_verbose = false) {
    if(!m_freader.is_open()) {
      m_freader.open(m_file.c_str());
      if(m_freader.fail()) {
        a_out << "inlib::rcsv::ntuple::initialize :"
              << " can't open " << m_file << "."
              << std::endl;
        return false;
      }
    }
    return ntuple::initialize(a_out,a_sep,a_suffix,a_verbose);
  }
protected:
  std::string m_file;
  std::ifstream m_freader;
};

}}

#endif