Chapter40

The lucid use of Humdrum depends on understanding how each of the tools works and how it is possible to connect the tools to perform particular tasks. The power and creativity of Humdrum truly lies in the hands of the user.

As we've seen, there are two parts to Humdrum: the Humdrum syntax and the Humdrum toolkit. The Humdrum syntax simply provides a formal framework for representing any kind of sequential symbolic data. A number of representation schemes are pre-defined in Humdrum, but you are free to construct your own representations as demanded by the tasks. In several of the tutorial examples in the previous chapters we saw various ad hoc representations that were concocted as temporary or intermediate representations. You should now feel comfortable with the possibilities of devising your own representations to assist you in whatever task you are interested.

The Humdrum Toolkit is a set of inter-related software tools. These tools manipulate text data conforming to the Humdrum syntax. If the data represents music-related information, then we can say that the Humdrum tools manipulate music-related information. Each Humdrum tool carries out a fairly modest process.

By way of review, we can group the various Humdrum (and UNIX) tools according to the type of operation. There are roughly a dozen or so classes of tasks:

• Displaying things. The ms command can be used to print or display musical notation.
• Auditing. MIDI sound output can be generated using the midi and perform commands.
• Searching for things. When searching for things within specified files, appropriate commands include: grep, egrep, yank -m, patt, pattern, correl, simil and humdrum -v. When searching for files that meet certain conditions, appropriate commands include grep -l, egrep -l and find. Most of these search tools rely extensively on regular expressions to define patterns of characters.
• Counting things. Appropriate commands include: wc, wc -l, grep -c, egrep -c, census and census -k. Eliminating unnecessary or confounding information can be achieved using rid, extract, grep -v, sed and humsed.
• Editing things. Manual editing may be done using any text editor, such as emacs or vi. Automated (or "stream") editing may be done using sed and humsed.
• Editorializing. E.g., add an editorial footnote to a specified note or passage; indicate that a passage differs from the composer's autograph.
• Transforming or translating between representations. Appropriate commands include: cents, deg, degree, dur, freq, hint, humsed, iv, kern, mint, pc, pf, pitch, reihe, semits, solfa, solfg, text, tonh, trans and vox.
• Arithmetic transformations of representations. Manipulating numerical values can be done using xdelta, ydelta, recode and awk.
• Extracting or selecting information. Appropriate commands include: extract, yank, grep, egrep, yank -m, thru, strophe, rend and csplit.
• Linking or joining information. Appropriate commands include: assemble, cat, cleave, timebase, context, ditto and join.
• Generating inventories of things. Appropriate commands include: sort and uniq. Once again, unnecessary or confounding information can be eliminated using rid, extract, grep -v, yank -m, sed or humsed.
• Classifying things. Numerical values can be classified using recode; non-numerical data can be classified using humsed.
• Labelling things. Appropriate commands include: patt -t, recode, humsed and timebase.
• Comparing whether things are the same or similar. Appropriate commands include: diff, diff3, cmp, correl and simil.
• Capturing data. MIDI data can be input via encode and record.
• Trouble-shooting. Appropriate commands include: humdrum, humdrum -v, proof, proof -k, veritas, midi and perform.

Not all of the existing Humdrum Tools were covered in this book. Nor were all of the available options described for all of the tools discussed. Exploring the Humdrum Reference Manual is recommended for readers interested in continuing to develop facility with Humdrum.

• What do I want my final output to look like?
  Do I want a count or inventory?

  115 instances of ...
28 instances of ...

  Do I want to output a found pattern?

  pattern found in line ...
pattern not found ...

  Do I want a comparison?

  file X is the same as file Y
X is similar to Y
X and Y are different

• What materials are available for processing?

  Use find and grep to locate useful materials.

• What materials do I need to create?

  Use encode to create new data. Use humdrum and proof to check the data. If necessary, define your own Humdrum representation for a given purpose.

• How do I transform my data so it is easier to process?

  Use recode and humsed to classify data into various classes -- such as up, down, leap, long, short, difficult, easy, clarion register, dominant, etc.

  Use translating/transforming commands such as mint, ydelta, pcset, etc to translate the data to a different representation.

• What data should I eliminate?

  Use rid, extract, yank, sed, humsed, uniq, uniq -d and grep -v to eliminate selective materials.

• What data do I need to coordinate?

  Use context to generate contextual information. Use assemble, rend and cleave to link information together.

• How do I know my results are worthwhile?

  Use comparative tests whenever you can. Use scramble -r, scramble -t, tac and reihe -s to generate control groups.

• Previous Chapter
• Table of Contents
• Detailed Contents
  
  © Copyright 1999 David Huron