Humdrum Extras

From CCARH Wiki
Jump to navigation Jump to search

Humdrum Extras is a set of command-line programs and C++ parser library for processing Humdrum files. The programs can be compiled for Linux, Apple OS X, or Windows (primarily within cygwin, but also in Visual C++). The Humdrum Extras library can be used to parse Humdrum files independent of the example programs provided with the package.


Example Programs

The primary intent of the Humdrum Extras package is for user-based processing of Humdrum files as an auxiliary to the Humdrum Toolkit. Since the programs are compiled from C++ code, they process data much faster than programs written in interpreted languages, such as AWK which is the main development language for the Humdrum Toolkit.

Documentation for example programs can be found on the web at extras.humdrum.org/man. The source code for these programs is found in the download file, within the src-programs directory, or they can be viewed online.




Programming Examples

Basic data access

humecho.cpp

Here is a very simple C++ program called humecho.cpp that uses the Humdrum file parser in the Humdrum Extras library:

    #include "humdrum.h"
    #include <iostream>
    using namespace std;
    
    int main(int argc, char** argv) {
       HumdrumFile hfile;
       if (argc > 1) hfile.read(argv[1]);
       else hfile.read(cin);
       cout << hfile;
       return 0;
    }
    


This program will take one Humdrum file as an argument (or standard input) and echo the contents of the Humdrum file to standard output. To compile this program using the Humdrum Extras makefiles, place humecho.cpp in the directory humextra/src-programs, and then type "make humecho" in the humextra directory. The compiled program will be placed in bin/humecho. The humecho program can be utilized in several ways, including downloading from the web, or using the humdrum:// URI (or hum:// or h:// abbreviations):

    cat file.krn | bin/humecho           | less     # standard input
    bin/humecho file.krn                 | less     # command-line argument
    bin/humecho h://wtc/wtc1f01.krn      | less     # humdrum:// URI
    bin/humecho http://y.z.com/file.krn  | less     # URL
    




humecho2.cpp (Accessing individual lines)

The humecho program shows how to access the datafile in its entirety. The following source code for humecho2.cpp demonstrates how to access lines in the file individually. A HumdrumFile class essentially consists of an array of HumdrumRecord classes, and HumdrumRecord classes essentially are character strings which print tab-delimited with cout:

    #include "humdrum.h"
    
    int main(int argc, char** argv) {
       HumdrumFile hfile;
       if (argc > 1) hfile.read(argv[1]);
       else hfile.read(std::cin);
       for (int i=0; i<hfile.getNumLines(); i++) {
          std::cout << hfile[i] << std::endl;
       }
       return 0;
    }
    

hfile.getNumLines() returns the number of text lines in the Humdrum file stored in the hfile variable. So the for loop iterates through each line in the file and prints it to standard output.




humecho3.cpp (Accessing spine data)

An even more verbose version of humecho is given below. The humecho3 program mimics the << operator for HumdrumRecords as a second for-loop. Each HumdrumRecord representing a line of music can be thought of as an array of strings (const char*), with each string being one token in the Humdrum file structure.

    #include "humdrum.h"
    
    int main(int argc, char** argv) {
       HumdrumFile hfile;
       if (argc > 1) hfile.read(argv[1]);
       else hfile.read(std::cin);
       for (int i=0; i<hfile.getNumLines(); i++) {
          std::cout << "\t" << hfile[i][0];
          for (int j=1; j<hfile[i].getFieldCount(); j++) {
             std::cout << "\t" << hfile[i][j] << std::endl;
          }
          std::cout << std::endl;
       }
       return 0;
    }
    

HumdrumRecords always contain at least one field, so the code "cout << hfile[i][0];" will not cause an invalid array access in any situation. Both [] operators used on the hfile variable (first to access a HumdrumRecord, and the second for a const char*) are checked for a valid range, and the program will exit with an error if an out-of-range value is requested.

The code hfile[i].getFieldCount() returns the number of "fields" on the line. This is a non-standard term for Humdrum files, since "spines" and "tokens" can have somewhat ambiguous meanings. The field count is a count of the spines, but if the spines split the count would include the subspines as well. Global comments and reference records are always element 0 in a HumdrumRecord line. Empty lines, which are technically not allowed in Humdrum files, are also acessed as an empty string at element 0.

Note that hfile[i][j] is a const char* and not a char*. If you want to change the contents of a field, you would have to use hfile[i].changeField(j, "new string").




HumdrumRecord line types

Each HumdrumRecord possesses an enumerated type:

    Enumeration Description
    E_humrec_data data lines other than measure
    E_humrec_data_measure line starting with “=”
    E_humrec_interpretation line starting with “*”
    E_humrec_bibliography reference records of the form “!!!key: value”
    E_humrec_global_comment   starts with “!!”, other than reference records
    E_humrec_local_comment local comment, starting with single "!"
    E_humrec_empty empty line


Use the HumdrumRecord::getType() function to access the type of a line. These type values can be used for switch statements, but for better code readability, the following helper HumdrumRecord functions interface with these enumerations:

    HumdrumRecord
    method
    Description
    .isData() true if data (other than barline).
    .isMeasure() true if barline (line starts with “=”).
    .isInterpretation() true if line starts with “*”.
    .isBibliographic() true if in the form of “!!!key: value”.
    .isGlobalComment()    true if line starts with “!!” and not bib.
    .isLocalComment() true if line starts with one “!”.
    .isEmpty() true if nothing on line.

In addition there are a few composite test for line types:

    HumdrumRecord
    method
    Description
    .isComment()    isBibliographic() or isGlobalComment() or isLocalComment()
    .isTandem() Interpretation lines which contain only spine manipulators (*+, *-, *^, *v, *x, or exclusive interpretations (starting with **).
    .isNull() isData() and all fields are "." (null token), or isInterpretation() and all fields are "*".
    .hasSpines() isData() or isMeasure() or isLocalComment() or isInterpretation()




"rid -GLI" (Remove all lines except for data lines)

The Humdrum Tool rid with the -GLI options can be implemented using the following C++ code:

    #include "humdrum.h"
    int main(int argc, char** argv) {
       HumdrumFile hfile(argv[1]);
       for (int i=0; i<hfile.getNumLines(); i++) {
          if (!(hfile[i].isData() || hfile[i].isMeasure())) continue;
          std::cout << hfile[i] << std::endl;
       }
       return 0;
    }
    

The above code will only print lines which are data or barlines. The official Humdrum file specification does not technically distinguish between barlines and data, but in practice and from a logical point of view they must be separated. So when using the Humdrum Extras C++ parser for Humdrum files, a line of data should not contain a mixture of data (or null tokens) and barlines.




"rid -GLId" (Remove comments, interpretations and null data)

    #include "humdrum.h"
    int main(int argc, char** argv) {
       HumdrumFile hfile(argv[1]);
       for (int i=0; i<hfile.getNumLines(); i++) {
          if (!(hfile[i].isData() || hfile[i].isMeasure())) continue;
          if (hfile[i].isNull()) continue;
          std::cout << hfile[i] << std::endl;
       }
       return 0;
    }
    

The HumdrumRecord::isNull() returns true if all fields in the record are equal to the string "." (called a null token in Humdrum terminology—not related to a NULL pointer in C).




User-specified Options

"myrid -M -C -I" (Handling command-line options)

The Humdrum Extras library contains a helper class called Options which can be used to manage command-line options. The following example program implements the options -M (suppress measure lines), -C (suppress comments), -I (suppress interpretations) in a C++ implementation of the Humdrum Toolkit rid program.

The Options class can be used to define multiple aliases for the same option, such as a short abbreviation and a long form. The options are formulated on the command line according to POSIX rules for options: single-letter options are preceded by a single dash. Multiple-letter options are preceeded by two dashes. When a single-letter option does not require it's own argument, they can be globbed together into a list of options preceded by a single dash. Here are various program usages for the code below:

    Command Description
    myrid -M file.krn Remove measure lines when echoing file.krn to standard output.
    myrid -M -I -C file.krn Remove measure lines, interpretations and comments (global, local and reference).
    myrid -MIC file.krn Same as above. Shorthand for bundling multiple single-letter boolean options.
    myrid --no-measures file.krn   Long form of "myrid -M".
    myrid --options Secret built-in option for the Option class which will force a list of defined options to be printed to standard output.
    myrid -A file.krn The option list will also be displayed when an undefined or misspelled option is used. Use "--" to disable options processing for unusual cases such as a filename starting with a dash.
    myrid -MM file.krn Duplicate options are ignored, so only the last -M is used. Note that this is not the option "MM" which would be formulated as "myrid --MM".
    myrid -M file.krn -IC Options can occur in any order, and can come before or after any command arguments which are not options.
    myride -M -- -file.krn -C Process the poorly named file "-file.krn" and the even more poorly named file called "-C" which is not an option if it comes after the -- marker.

Note in the following source code, an extra include directive for the Options class does not need to be added, since the declaration for the Options class is included in humdrum.h. If you want to use the Options class independent of the HumdrumFile parser, you can instead include the file "Options.h".

    #include "humdrum.h"
    int main(int argc, char** argv) {
       Options opts;
       opts.define("M|no-measures:b", "remove measures");
       opts.define("C|no-comments:b", "remove comments");
       opts.define("I|no-interpretations:b", "remove interpretations");
       opts.process(argc, argv);
       int measuresQ = !opts.getBoolean("no-measures");
       int commentsQ = !opts.getBoolean("no-comments");
       int interpQ = !opts.getBoolean("no-interpretations");
       HumdrumFile hfile(opts.getArg(1));
       for (int i=0; i<hfile.getNumLines(); i++) {
          if (hfile[i].isMeasure() && !measureQ) continue;
          if (hfile[i].isComment() && !commentQ) continue;
          if (hfile[i].isInterpretation() && !interpQ) continue;
          std::cout << hfile[i] << std::endl;
       }
       return 0;
    }
    

The code "HumdrumFile hfile(opts.getArg(1));" reads data from the first argument on the command line. Note that argument counts are indexed from 1 rather than 0. Perhaps not a great thing to do, but was intended to allow for similar behavior with command-line string arrays in C, where the name of the command is stored in array element 0, and the first argument (or option) is stored in array element 1. To access the name of the command, use the Options::getCommand() function.




Option definitions

Notice the Options::define() function calls in the above program. These are used to define the options that an Options variable will search for when the Options::process() function is called. The .define() function takes two arguments (the second one optional). The first argument is the definition string, and the second is a human-readable description of the option.

The option definition string has the basic format:

"OptionName=OptionType:DefaultValue"

The OptionName can include aliases which are added to the Option name, separated by a pipe (|) character:

"OptionName|OptionAlias1|OptionAlias2=OptionType:DefaultValue"

For example:

"M|no-measures=b"

Is the definition of the option "M" or equivalently "no-measures" which is a boolean type (which means that it sets a true/false switch for the option). For boolean options, there is no default value—they are "false" if not given as an argument to the program, and turned to "true" when given as input to a program.

There are four Options data types:

    Option type Description Options value access
    b boolean (true or false) .getBoolean("OptionName")
    i integer .getInteger("OptionName")
    d double (floating-point number) .getDouble("OptionName")
    s string .getString("OptionName")

In terms of implementation, there are really only two types: booleans (with out parameters) and non-booleans (with parameters). Within a C++ program you can acess the original string form of the option's parameter, or you can convert it into an int or a double at runtime. For example, if an option "number" is defined, you can get the integer version of the number with .getInteger("number"), or the double version of the number with .getDouble("number"), or you can check to see if the option was set from the input arguments to the program with .getBoolean("number").

Here are some example option definitions with option names, option aliases, and option types:

    Option definition Command-line examples
    "r=b" command -r
    "m=i" command -m 10 or command -m10
    value=d" command -v 5.23 or command -v5.23
    command --value 5.23
    command --value=5.23
    "t=s" command -t string or command -tstring
    command -t "string with spaces"
    command -t 'funny $tring'

When options names (or option aliases) are a single character, the space between the option name and it parameter is optional, as in "command -m 10" or "command -m10". When an option has multiple characters, the space is not optional, although an equals sign can be substituted for the space: "command --value 5.23" and "command --value=5.23". When a string option contains spaces, or other special characters reserved for shell syntax, (such as [;&$|?*\]). The multi-word option must be enclosed in quotes. To insert a quote into the string option place a backslash before it: \". To prevent the command-line parser from looking inside of the string use single quotes: "command -t 'funny $tring'". In this case the final input will be "funny $tring". If double quotes were used, $tring would be interpreted as an environmental variable and its value would be substituted, usually resulting in "funny ", since you are not likely to have the shell variable $tring defined.




Default option values

The final component of the option definition is a default value to use if no input is given for that option on the command-line. If no default value is given in the definition, the default value will be zero. For example, if this option definition is given:

       options.define("v|val|value=i:10", "an integer value");
    

Then here are different behaviors when accessing that option's value in C++:

    User-set option: 
       program -v 20
          options.getInteger("value")      → 20
          options.getInteger("val")        → 20
          options.getInteger("v")          → 20
    
    Default option:
       program
          options.getInteger("value")      → 10
          options.getInteger("val")        → 10
          options.getInteger("v")          → 10
    




Accessing option values

As mentioned previously, the .getBoolean, .getInteger, .getDouble and .getString accessor functions are used to extract an option value from the Options database after .process() has been called on the argc and argv input parameters to main(). All of the get functions can be applied to any option type. For example, using the option definition:

       .define("t|tempature=d:80.6 Farenheit", "temperature setting")
    

can be used to extract any of the four option types in C++:

       .getBoolean("temperature")           → 1 (true) if set via the command-line.
                                            → 0 (false) if not set via the command-line.
       .getInteger("temperature")           → 80
       .getDouble("temperature")            → 80.6
       .getString("temperature")            → "80.6 Farenheit"
    




Input from piped data or file(s)

Most of the previous program examples expect a single filename as input for processing. The following program example (humecho4 is more flexible, allowing for multiple input files. If no filenames are given, then standard input will be read as the input data:

    #include "humdrum.h"
    using namespace std;
    int main(int argc, char** argv) {
       Options options(argc, argv);
       options.process();
       HumdrumFile hfile;
       int numinputs = options.getArgCount();
       for (int i=1; i<=numinputs || i==0; i++) {
          if (numinputs < 1) {
             hfile.read(cin); // read from standard input
          } else {
             hfile.read(options.getArg(i));
          }
          // do something with the Humdrum data here:
          cout << hfile;
       }
       return 0;
    }
    

This program has an identical function to humecho.cpp, but now multiple files can be read in and processed at the same time. For example if there are two input files with these contents:


    file 1
    **kern
    1c
    2d
    4e
    *-
    file 2
    **kern
    2cc
    4b
    2a
    *-
    output
    **kern
    1c
    2d
    4e
    *-
    **kern
    2cc
    4b
    2a
    *-

Here are some possible command-line realizations for the above program:

    humecho4 file.krn
    humecho4 file1.krn file2.krn file3.krn
    cat file.krn | humecho4
    humecho4
    

The last command will cause the shell to wait while you type in the input to humecho4, followed by control-D to indicate the end of input data.

Note that the number of command-line arguments (other than options) can be queried from an Options variable by using the .getArgCount() function. If there are three filenames as in "echo4 file1.krn file2.krn file3.krn", then .getArgCount() will return 3. The .getArg() function will return a string for the specified argument, starting with argument 1: .getArg(1) == file1.krn, .getArg(2) == file2.krn, .getArg(3) == file3.krn. Note that the first argument is not .getArg(0). If you want to access the command name, then use .getCommand(), which would return "humecho4" in this case.

When reading from standard input use HumdrumFile::read(istream) rather than HumdrumFile::read(const char*). For example, reading from standard input is done with hfile.read(cin) in the above code.




C string comparison functions

Here are three of the string comparison functions available within in the C (or C++) language:

strcmp("string1", "string2")
returns 0 if strings are equivalent
returns –1 if string1 is alphabetized before string2
returns +1 if string1 is alphabetized after string2.
strncmp("string1", "string2", n)
compare only first n characters of the two strings.
strchr("string", 'character')
returns a pointer to the first occurrence of the character within the string. If the character is not found in the string, returns a NULL pointer.

Other interesting string processing functions in the C language are strstr which is similar to strchr but search for a sub-string within the a string; and strrchr which is similar to strchr but searches for the character in the reverse direction in the string, which returns the last occurrence of the character in the string (or NULL) if the character is not in the string. For more description about these functions, type "man strrchr" in a terminal for more information about the strrchr function (or any other standard C fuction).




Third dimension of data access (Note-level access)

Accessing individual notes in **kern data spines requires three dimensions of indexing: (1) the data line of the note, the data field on the line for the note, and then the note number within a chord for the note. Previous program examples demonstrated how to access lines and line-fields. The following program (noteloc) goes one step further to access individual **kern notes. The program takes any sort of Humdrum file, and then outputs a list of all notes found in all kern spines:

    #include "humdrum.h"
    int main(int argc, char** argv) {
       Options options(argc, argv);
       options.process();
       HumdrumFile hfile;
       hfile.read(options.getArg(1));
       char buffer[1024] = {0};
       for (int i=0; i<hfile.getNumLines(); i++) {
          if (!hfile[i].isData()) continue; // ignore non-data lines
          for (int j=0; j<hfile[i].getFieldCount(); j++) {
             if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
             if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
             int count = hfile[i].getTokenCount(j);
             for (int k=0; k<count; k++) {
                cout << "(" << i+1 <<"," << j+1 << "," << k+1 << ")\t"
                     << hfile[i].getToken(buffer, j, k) << endl;
             } 
          }
       }
       return 0;
    }
    

The line:

if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;

is used to skip over all spines which do not have **kern data. The function .getExInterp() returns a const char* string for the name of the exclusive interpretation. The strcmp() function compares the returns exclusive interpretation name with the string "**kern", and if it does not match, the next data field on the line will be examined. An equivalent way of identifying the exclusive interpretation can be done with the .isExInterp() function. The following line of code is equivalent to the one above:

if (hfile[i].isExInterp(j, "**kern")) continue;

If the input to the program is the following:

**kern	**text	**kern
4C	ig-	4c
4D 4E	-no-	.
4F	-red	.
.	.	4d 4e
4r	.	.
4G 4A 4B	text	.
*-	*-	*-


Then the output from the noteloc program will be:


  (2,1,1) 4C
  (2,3,1) 4c
  (3,1,1) 4D
  (3,1,2) 4E
  (4,1,1) 4F
  (5,3,1) 4d
  (5,3,2) 4e
  (6,1,1) 4r
  (7,1,1) 4G
  (7,1,2) 4A
  (7,1,3) 4B

Each of the three numbers before the note indicates the address within the file for the note, with the first number being the line on which the note occurs, the second number the field on the line which contains the note, and the last number is the note number within the (possible) chord for the note.




kerninfo.cpp (Count **kern notes in data)

Here is an example program which somewhat emulates the "census -k" command from the Humdrum Toolkit. The program will count the number of note attacks, rests and tied notes in one or more Humdrum files.

    #include "humdrum.h"
    using namespace std;
    int main(int argc, char** argv) {
       Options options(argc, argv);
       options.process();
       HumdrumFile hfile;
       int restcount   = 0;
       int nullcount   = 0;
       int attackcount = 0;
       int tiedcount   = 0;
       int chordcount  = 0;
       for (int arg=1; arg <= options.getArgCount() || arg == 0; arg++) {
          if (options.getArgCount() == 0) {  hfile.read(cin); } 
          else { hfile.read(options.getArg(arg)); }
          char buffer[1024] = {0};
          for (int i=0; i<hfile.getNumLines(); i++) {
             if (!hfile[i].isData()) continue;
             for (int j=0; j<hfile[i].getFieldCount(); j++) {
                if (!hfile[i].isExInterp(j, "**kern")) continue;
                int count = hfile[i].getTokenCount(j);
                if (count > 1) chordcount++;
                for (int k=0; k<count; k++) {
                   hfile[i].getToken(buffer, j, k);
                   if (strchr(buffer, 'r') != NULL)   { restcount++; } 
                   else if (strcmp(buffer, ".") == 0) { nullcount++; } 
                   else if (strchr(buffer, '_') != NULL) { /* ignore */ }
                   else if (strchr(buffer, ']') != NULL) { tiedcount++; } 
                   else { attackcount++; }
                }
             }   
          }
       }
       cout << "Note attacks: " << attackcount << endl;   
       cout << "Tied notes  : " << tiedcount   << endl;
       cout << "Chords      : " << chordcount  << endl;   
       cout << "Rests       : " << restcount   << endl;
       cout << "Null Tokens : " << nullcount   << endl;   
       return 0;
    }
    
    Trying out the kerninfo prorgram on this input data:
    **kern	**text	**kern
    4C	ig-	4c
    4D 4E	-no-	.
    4F	-red	.
    .	.	4d 4e
    4r	.	.
    4G 4A 4B	text	.
    *-	*-	*-

    Results in these statistics:

    Note attacks: 10
    Tied notes  : 0
    Chords      : 3
    Rests       : 1
    Null Tokens : 5
    

Trying out the kerninfo program on a real piece of music:

  kerninfo h://wtc/wtc1p04.krn
  Note attacks: 675
  Tied notes  : 85
  Chords      : 14
  Rests       : 69
  Null Tokens : 967





Convert class

In addition to the Options class, and important helper class in HumdrumExtras is the Convert class. This class handles most conversions between data types. The HumdrumFile class essentially stores a two-dimensional array of strings. The **kern notes in a HumdrumFile variable are extracted as strings, but will need to be interpreted further depending on the information about the note which you need. For example, to convert a **kern note into a MIDI note number, use the following Convert function:

  Convert::kernToMidiNoteNumber("4d-")          →  61

Likewise, the MIDI note 61 can be converted back into a **kern note:

  Convert::midiNoteNumberToKern(buffer, 61)     →  "c#"

All access to Convert class functions is done statically, so you can shorten the code by using the a typedef for Convert to a shorter name:

   typedef Convert C;
   C::kernToMidiNoteNumber("4d-");




Convert **kern note name to MIDI

The following program will convert the first note of every chord into a MIDI note number. As an exercise, adjust the code so that it prints a MIDI note number for every note in the chords.

#include "humdrum.h"
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile(options.getArg(1));
   for (int i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue;
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
      if (hfile[i].isExInterp(j, "**kern")) continue;
      if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
      if (strchr(hfile[i][j], 'r') != NULL) continue; // ignore rests
         cout << hfile[i][j] << "\t" << Convert::kernToMidiNoteNumber(hfile[i][j]) << endl;
      }
   }
   return 0;
}

Example input and output:

**kern	**text	**kern
4C	ig-	4c
4D 4E	-no-	.
4F	-red	.
.	.	4d 4e
4r	.	.
4G 4A 4B	text	.
*-	*-	*-
  4C        48
  4c        60
  4D 4E     50
  4F        53
  4d 4e     62
  4G 4A 4B  55





Note Histogram (notehist.cpp)

Here is an example of how to count the number of twelvetone pitch classes in a Humdrum file. The following program will count tied notes. As an exercise, have the program skip counting of any middle or end tied notes (middle tied notes have an underscore (_) in their content, and ending tied notes has a closing square bracket (])).

#include "humdrum.h"
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   double histogram[12] = {0};
   char buffer[1024] = {0};
   int midikey;
   int i;
   for (i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue; // ignore non-data lines
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
         if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
         if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
         int count = hfile[i].getTokenCount(j);
         for (int k=0; k<count; k++) {
            hfile[i].getToken(buffer, j, k);
            if (strchr(buffer, 'r') != NULL) continue; // ignore rests
            midikey = Convert::kernToMidiNoteNumber(buffer);
            histogram[midikey % 12]++;
         }
      }
   }
   for (i=0; i<12; i++) {
      std::cout << i << "\t" << histogram[i] << std::endl;
   }
   return 0;
}

Example output when processing a real piece of music is given below. The first output line means there are 600 C notes in Beethoven's 32nd sonata, mvmt. 1, 233 Cs/Ds, etc.

  notehist h://beethoven/sonatas/sonata32-1.krn
  0	600 
  1	233
  2	279
  3	476
  4	146
  5	513
  6	144
  7	636
  8	459
  9	121
  10	259
  11	230

To sort the pitch classes by how often they occur:

  notehist h://beethoven/sonatas/sonata32-1.krn | sort -nrk2
  7	636
  0	600
  5	513
  3	476
  8	459
  2	279
  10	259
  1	233
  11	230 
  4	146
  6	144
  9	121

In this case the most common pitch class is G (7), and the least common pitch class is A (9). Note that the sorting did not have to be implemented in the C++ program, since the command-line program sort could be utilized. The options to sort are -n (sort numerically), -r (sort in reverse order so that the largest number comes first), and -k2 (sort by the second column of data).

Here is a modification of the program so that pitch names rather than pitch-class numbers are displayed:

#include "humdrum.h"
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   double histogram[12] = {0};
   char buffer[1024] = {0};
   int midikey;
   int i;
   for (i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue; // ignore non-data lines
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
         if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
         if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
         int count = hfile[i].getTokenCount(j);
         for (int k=0; k<count; k++) {
            hfile[i].getToken(buffer, j, k);
            if (strchr(buffer, 'r') != NULL) continue; // ignore rests
            midikey = Convert::kernToMidiNoteNumber(buffer);
            histogram[midikey % 12]++;
         }
      }
   }
   for (i=0; i<12; i++) {
      std::cout << i << "\t" 
                << Convert::base12ToKern(buffer, histogram[i] + 4 * 12) << std::endl;
   }
   return 0;
}

So now the note names will be printed instead of their numeric equivalent:

  notehist h://beethoven/sonatas/sonata32-1.krn | sort -nrk2
  G	636
  C	600
  F	513
  E-	476
  G#	459
  D	279
  B-	259
  C#	233  
  B	230 
  E	146
  F#	144 
  A	121

If you want to preserve the accidental spellings, then you can use base-40 instead of base-12 (MIDI note numbers):

#include "humdrum.h"
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   double histogram[40] = {0};
   char buffer[1024] = {0};
   int base40;
   int i;
   for (i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue; // ignore non-data lines
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
         if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
         if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
         int count = hfile[i].getTokenCount(j);
         for (int k=0; k<count; k++) {
            hfile[i].getToken(buffer, j, k);
            if (strchr(buffer, 'r') != NULL) continue; // ignore rests
            base40 = Convert::kernToBase40(buffer);
            histogram[base40 % 40]++;
         }
      }
   }
   for (i=0; i<40; i++) {
      if (histogram[i] == 0) { continue; }
      std::cout << Convert::base40ToKern(buffer, i + 3*40) << "\t"
                << histogram[i] << std::endl;
   }
   return 0;
}

With the more verbose pitch information being:

notehist h://beethoven/sonatas/sonata32-1.krn

  C-	32
  C	600
  C#	1
  D-	232
  D	271
  D#	1
  E--	8
  E-	475
  E	134
  E#	2
  F-	12
  F	511
  F#	83
  G-	61
  G	636
  G#	2
  A-	457
  A	121
  B-	259
  B	198


Standard Template Library classes such as vector can be used instead of the C-like histogram array used in the previous program. Here is an example using the vector class:

#include "humdrum.h"
#include <vector>
using namespace std;

int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   vector<int> histogram(40);
   char buffer[1024] = {0};
   int base40;
   unsigned int i;
   for (i=0; i<(unsigned int)hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue; // ignore non-data lines
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
         if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
         if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
         int count = hfile[i].getTokenCount(j);
         for (int k=0; k<count; k++) {
            hfile[i].getToken(buffer, j, k);
            if (strchr(buffer, 'r') != NULL) continue; // ignore rests
            base40 = Convert::kernToBase40(buffer);
            histogram[base40 % 40]++;
         }
      }
   }
   for (i=0; i<histogram.size(); i++) {
      if (histogram[i] == 0) { continue; }
      cout << Convert::base40ToKern(buffer, i + 3*40) << "\t"
                << histogram[i] << endl;
   }
   return 0;
}

In addition, Humdrum Extras has an Array class which has a similar functionality as the vector class or C arrays:

#include "humdrum.h"
using namespace std;

int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   Array<int> histogram(40);
   histogram.setAll(0);
   char buffer[1024] = {0};
   int base40;
   int i;
   for (i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue; // ignore non-data lines
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
         if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
         if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
         int count = hfile[i].getTokenCount(j);
         for (int k=0; k<count; k++) {
            hfile[i].getToken(buffer, j, k);
            if (strchr(buffer, 'r') != NULL) continue; // ignore rests
            base40 = Convert::kernToBase40(buffer);
            histogram[base40 % 40]++;
         }
      }
   }
   for (i=0; i<histogram.getSize(); i++) {
      if (histogram[i] == 0) { continue; }
      cout << Convert::base40ToKern(buffer, i + 3*40) << "\t"
                << histogram[i] << endl;
   }
   return 0;
}

Duration-weighted note histogram

Often it is useful to know how long a certain pitch class sounds in a musical work rather than just how many note attacks for each pitch class. Here is a program which measures the duration of each pitch class in the music:

#include "humdrum.h"
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   Array<double> histogram(12);
   histogram.setAll(0);
   histogram.allowGrowth(0);
   char buffer[1024] = {0};
   double duration;
   int midikey;
   int i;
   for (i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) continue; // ignore non-data lines
      for (int j=0; j<hfile[i].getFieldCount(); j++) {
         if (strcmp("**kern", hfile[i].getExInterp(j)) != 0) continue;
         if (strcmp(".", hfile[i][j]) == 0) continue; // ignore null tokens
         int count = hfile[i].getTokenCount(j);
         for (int k=0; k<count; k++) {
            hfile[i].getToken(buffer, j, k);
            if (strchr(buffer, 'r') != NULL) continue; // ignore rests
            midikey = Convert::kernToMidiNoteNumber(buffer);
            duration = Convert::kernToDuration(buffer);
            histogram[midikey % 12] += duration;
         }
      }
   }
   for (i=0; i<histogram.getSize(); i++) {
      std::cout << Convert::base12ToKern(buffer, i+4*12) << "\t"
                << histogram[i] << std::endl;
   }
   return 0;
}


Now the program will output the duration in quarter notes for each pitch class in the music:

  C	319.556
  C#	98.6833 
  D	137.644
  E-	208.4
  E	84.1444
  F	225.158
  F#	75.5278
  G	308.772
  G#	203.953
  A	61.2111
  B-	121.267
  B	125.767




Line field enumeration by spine

The HumdrumRecord::getFieldCount() function returns the number of tokens on each Humdrum line, and each of these fields is accessed by an index number up to this count. However, if you want to access data by spine number (or to be clear primary spine number), you have to use a different method. The field number and spine number do not always match because spines can split into subspines which will increase the field count on a line.

To access data by spine, use the HumdrumRecord::getPrimaryTrack() function. This function returns an integer value for the current spine, enumerated from one (sorry). Multiple fields can have the same primary track number, which is caused by a spine split in the data.

Here is a program which prints the primary track for each data token in a Humdrum file structure:

#include "humdrum.h"
using namespace std;
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   for (int i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) {
         cout << hfile[i] << endl;
         continue;
      }
      cout << hfile[i].getPrimaryTrack(0);
      for (int j=1; j<hfile[i].getFieldCount(); j++) {
         cout << '\t' << hfile[i].getPrimaryTrack(j);
      }
      cout << endl;
   }
   return 0;
}

And here is example input and output:

**a	**b	**c
.	.	.
.	.	.
*	*^	*^
.	.	.	.	.
.	.	.	.	.
*	*v	*v	*	*
*	*	*v	*v
.	.	.
.	.	.
*-	*-	*-
**a	**b	**c
1	2	3
1	2	3
*	*^	*^
1	2	2	3	3
1	2	2	3	3
*	*v	*v	*	*
*	*	*v	*v
1	2	3
1	2	3
*-	*-	*-

Notice that the input data contains three exclusive interpretations. This will mean that there are three primary tracks (or spines) in the data. In the output you can see the numbers for each track: 1, 2, and 3. The HumdrumRecord::getMaxTrack() function can be called to find out what the maximum track number is (3 in this case).

The primary track value is an integer. If you also want to see the subtrack information, use .getTrack() instead of .getPrimaryTrack(). The .getTrack() function will have the primary track in the integer portion of a double float, and the enumerated subtrack in the fractional value. The subtrack value uses the first three digits of the fraction, so you can extract the subtrack by removing the integer part of the number and multiplying by 1000. Note that the subtrack enumeration starts at 0, while the primary tracks are enumerated from 1 (sorry again).

#include "humdrum.h"
using namespace std;
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   for (int i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) {
         cout << hfile[i] << endl;
         continue;
      }
      cout << hfile[i].getPrimaryTrack(0);
      for (int j=1; j<hfile[i].getFieldCount(); j++) {
         cout << '\t' << hfile[i].getTrack(j);
      }
      cout << endl;
   }
   return 0;
}
**a	**b	**c
1	2	3
1	2	3
*	*^	*^
1	2	2.001	3	3.001
1	2	2.001	3	3.001
*	*v	*v	*	*
*	*	*v	*v
1	2	3
1	2	3
*-	*-	*-

Even more detail about track information can be accessed with the .getSpineInfo() function. This function returns the internally stored string which keeps track of how the spine/subspine was manipulated on previous lines of the data.

#include "humdrum.h"
using namespace std;
int main(int argc, char** argv) {
   Options options(argc, argv);
   options.process();
   HumdrumFile hfile;
   hfile.read(options.getArg(1));
   for (int i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].isData()) {
         cout << hfile[i] << endl;
         continue;
      }
      cout << hfile[i].getPrimaryTrack(0);
      for (int j=1; j<hfile[i].getFieldCount(); j++) {
         cout << '\t' << hfile[i].getSpineInfo(j);
      }
      cout << endl;
   }
   return 0;
}
**a	**b	**c
1	2	3
1	2	3
*	*^	*^
1	(2)a	(2)b	(3)a	(3)b
1	(2)a	(2)b	(3)a	(3)b
*	*v	*v	*	*
*	*	*v	*v
1	2	3
1	2	3
*-	*-	*-

In this case the spine info "(2)a" means that that token is in primary spine 2, but the spine was split once, and this subtrack is the left-hand subspine coming out of the split. The .getPrimaryTrack() function returns the first number in the .getSpineInfo() string.

Here is a more complex spine manipulation:

**a	**b	**c
.	.	.
*	*^	*^
.	.	.	.	.
*	*	*^	*	*
.	.	.	.	.	.
*	*	*v	*v	*	*
*	*v	*v	*	*
.	.	.	.
*	*	*v	*v
.	.	.
*-	*-	*-

Which has this spine information:

**a	**b	**c
1	2	3
*	*^	*^
1	(2)a	(2)b	(3)a	(3)b
*	*	*^	*	*
1	(2)a	((2)b)a	((2)b)b	(3)a	(3)b
*	*	*v	*v	*	*
*	*v	*v	*	*
1	2	(3)a	(3)b
*	*	*v	*v
1	2	3
*-	*-	*-




Spine manipulator examples

Here is an example of spine splits (*v) and joins (*v):

**a	**b	**c
.	.	.
*	*^	*
.	.	.	.
*	*	*^	*
.	.	.	.	.
*	*v	*v	*v	*
.	.	.
*-	*-	*-
**a	**b	**c
1	2	3
*	*^	*
1	(2)a	(2)b	3
*	*	*^	*
1	(2)a	((2)b)a	((2)b)b	3
*	*v	*v	*v	*
1	2	3
*-	*-	*-

Here is an example of spine additions (*+) and terminations (*-):

**a	**b
.	.
*
*+	**c
.	.	.
*	*-	*
.	.
*-	*-
**a	**b
1	2
*	*+	**c
1	2	3
*	*-	*
1	3
*-	*-

Here is an example of spine exchanges (*x):

**a	**b
.	.
.	.
*x	*x
.	.
.	.
*-	*-
**a	**b
1	2
1	2
*x	*x
2	1
2	1
*-	*-

And finally a complex example using all of the spine manipulators:

**a	**b
.	.
*	*^
.	.	.
*+	**c	*
.	.	.	.
*	*	*x	*x
.	.	.	.
*	*	*^	*
.	.	.	.	.
*	*+	**d	*	*
.	.	.	.	.	.
*	*	*	*	*x	*x
.	.	.	.	.	.
*	*-	*	*	*	*
.	.	.	.	.
*v	*v	*	*	*
.	.	.	.
*	*v	*v	*
.	.	.
*	*v	*v
.	.
*-	*-


**a     **b
1       2
*       *^
1       (2)a    (2)b
*+      **c     *
1       3       (2)a    (2)b
*       *       *x      *x
1       3       (2)b    (2)a
*       *       *^      *
1       3       ((2)b)a ((2)b)b (2)a
*       *+      **d     *       *
1       3       4       ((2)b)a ((2)b)b (2)a
*       *       *       *       *x      *x
1       3       4       ((2)b)a (2)a    ((2)b)b
*       *-      *       *       *       *
1       4       ((2)b)a (2)a    ((2)b)b
*v      *v      *       *       *
1       4       ((2)b)a (2)a    ((2)b)b
*       *v      *v      *
1       4       ((2)b)a (2)a    ((2)b)b
*       *v      *v
1       4       2
*-      *-





myextract.cpp

Now that you know how to extract information about spines and subspines, you can write your own version of the Humdrum Toolkit's extract program. It will be even more powerful than the extract program, since the extract program cannot deal with subspines without special processing. Here is the myextract code:

#include "humdrum.h"
using namespace std;

void extract(HumdrumFile& hfile, int primarytrack) {
   int i, j, fcount, pcount;
   for (i=0; i<hfile.getNumLines(); i++) {
      if (!hfile[i].hasSpines()) {
         cout << hfile[i] << endl;
         continue;
      }
      fcount = hfile[i].getFieldCount();
      pcount = 0;
      for (j=0; j<fcount; j++) {
         if (primarytrack == hfile[i].getPrimaryTrack(j)) {
            if (pcount++ > 0) cout << '\t';
            cout << hfile[i][j];
         }
      }
      if (pcount > 0) cout << endl;
   }
}

int main (int argc, char** argv) {
   Options opts;
   opts.define("t|track=i:1", "extract specified track");
   opts.process(argc, argv);
   int primarytrack = opts.getInteger("track");
   int numinputs = opts.getArgCount();
   HumdrumFile hfile;
   for (int i=1; i<=numinputs || i==0; i++) {
      if (numinputs < 1) { hfile.read(std::cin); }
      else { hfile.read(opts.getArg(i)); }
      extract(hfile, primarytrack);
   }
   return 0;
}

Here is input and output to the command:

   myextract -t 2
**a	**b	**c
a	b	c
*	*^	*
a	b1	b2	c
*	*v	*v	*
a	b	c
*-	*-	*-
**b
b
*^
b1	b2
*v	*v
b
*-

Notice that all of the "B" information which was in spine 2 was extracted from the input data.

Regular Expressions in C

Regular expressions are an important concept to understand when working with Humdrum data, since the data format was designed to take advantage of them.

GNU POSIX regular expressions

Different Operating systems have different C implementations of regular expressions. Here is an example of how to use them on most linux operating systems using the GNU POSIX regular expressions:

#include <regex.h>
#include <stdlib.h>
#include <stdio.h>
using namespace std;
int main(int argc, char** argv) {
   if (argc < 3) exit(1);
   const char *searchstring = argv[1];
   const char *datastring = argv[2];
   regex_t re;
   int flags = 0 | REG_EXTENDED | REG_ICASE;
   int status = regcomp(&re, searchstring, flags);
   if (status !=0) {
      char errstring[999];
      regerror(status, &re, errstring, 999);
      printf("%s\n", errstring);
      exit(1);
   }
   status = regexec(&re, datastring, 0, NULL, 0);
   if (status == 0) printf("Match Found\n");
   else printf("Match Not Found\n");
   return 0;
}

Example behavior of the program:

   search cat "cat in the hat"
   Match Found
  
   search dog "cat in the hat"
   Match Not Found

For the simple examples above the strstr() C library function could have been used (which would probably also run faster). But using regular expressions allows for more powerful generalized searching, such as looking for upper and lower case matches:

   search cat "Cat in the hat"
   Match Found

In this case "cat" was matched to "Cat" since the REG_ICASE flag was used to ignore difference between upper and lower letter cases. The REG_EXTENDED flag is for using extended regular expressions (regular expressions 2.0). The regexec() function returns 0 if a match was found, otherwise returns a non-zero value to indicate no match was found.


mysed.c (Search and replace)

The following program demonstrates how to do search and replace on strings with GNU POSIX regular expresions.

#include <regex.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
int main(int argc, char** argv) {
   if (argc < 4) exit(1);
   char buffer[1024] = {0};
   const char *searchstring = argv[1];
   const char *replacestring = argv[2];
   const char *datastring = argv[3];
   regex_t re;
   regmatch_t match;
   int compflags = 0 | REG_EXTENDED | REG_ICASE;
   int status = regcomp(&re, searchstring, compflags);
   if (status !=0) {
      regerror(status, &re, buffer, 1024);
      printf("%s\n", buffer);
      exit(1);
   }
   status = regexec(&re, datastring, 1, &match , 0);
   while (status == 0) {
      strncat(buffer, datastring, match.rm_so);
      strcat(buffer, replacestring);
      datastring += match.rm_eo;
      status = regexec(&re, datastring, 1, &match, REG_NOTBOL);
   }
   printf("%s%s\n", buffer, datastring);
   return 0;
}

Example use:

  mysed klm 000 abcdefghijklmnopqrstuvwxyz
  abcdefghij000nopqrstuvwxyz