[lilypond.git] / Documentation / metadoc / hacking.texi

\input texinfo @c -*-texinfo-*-
@setfilename internals.info
@settitle LilyPond internals

@node Top, LilyPond internals, (dir), (dir)
@top


@menu
* LilyPond internals::
* Overview::
* mudela::                      
* Request_engraver::            
* Graphic elements::
* Score elements::
* Items::
* Spanners::
* Future work::
* Coding standards::
* Making patches::

@end menu

@node LilyPond internals,  , Top, Top
@menu
* Overview::                      Overview
* mudela::                        mudela
* Request_engraver::              Request_engraver
@end menu


@chapter Help Needed!

[Call for help]

[List tasks:

* Mutopia

* Doco LilyPond

* Straightforward extensions

* GUI editors.

]

@chapter LilyPond internals


This documents some aspects of the internals of GNU LilyPond. Some of
this stuff comes from e-mail I wrote, some from e-mail others wrote,
some are large comments taken away from the headers. This page may be
a little incoherent.  Unfortunately, it is also quite outdated.  A
more thorough  and understandable document is in the works.

You should use @code{doc++} to take a peek at the sources.

@node Overview, mudela, Top, Top
@section Overview

GNU LilyPond is a "multi-pass" system. The different passes have been
created so that they do not depend on each other. In a later stage
some parts may be moved into libraries, or seperate programs, or they
might be integrated in larger systems.

@table @samp

@item Parsing:

No difficult algorithms. The .ly file is read, and converted to a list
of @code{Scores}, which each contain @code{Music} and paper/midi-definitions.

@item Interpreting music

The music is walked through in time-order. The iterators which do the
walking report Music to Translators which use this information to
create elements, either MIDI or "visual" elements. The translators
form a hierarchy; the ones for paper output are Engravers, for MIDI
Performers.

The translators swallow Music (mostly atomic gobs called Requests),
create elements, broadcast them to other translators on higher or same
level in the hierarchy:

The stem of a voice A is broadcast to the staff which contains A, but
not to the stems, beams and noteheads of a different voice (say B) or
a different staff. The stem and noteheads of A are coupled, because
the the Note_heads_engraver broadcasts its heads, and the Stem_engraver catches
these.

The engraver which agrees to handle a request decides whether to to
honor the request, ignore it, or merge it with other requests. Merging
of requests is preferably done with other requests done by members of
the same voicegroups (beams, brackets, stems). In this way you can put
the voices of 2 instruments in a conductor's score so they make chords
(the Beam requests of both instruments will be merged).

@item Prebreaking

Breakable stuff (eg. clefs and bars) are copied into pre and
postbreaks.

@item Preprocessing

Some dependencies are resolved, such as the direction of stems, beams,
and "horizontal" placement issues (the order of clefs,  keys etc,
placement of chords in multi-voice music), 

@item Break calculation:

The lines and horizontal positions of the columns are determined.

@item Breaking

Through some magical interactions with Line_of_score and Super_elem
(check out the source) the "lines" are produced.

All other spanners can figure across which lines they are spread. If
applicable, they break themselves into pieces. After this, each piece
(or, if there are no pieces, the original spanner itself) throws out
any dependencies which are in the wrong line.

@item Postprocesing:

Some items and all spanners need computation after the Paper_column
positions are determined. Examples: slurs, vertical positions of
staffs.

@item Output paper

@end table

@node mudela, Request_engraver, Overview, Top
@section mudela

Most information is stored in the form of a request.  In music
typesetting, the user might want to cram a lot more symbols on the
paper than actually fits. To reflect this idea (the user asks more
than we can do), the container for this data is called Request.

In a lot of other formats this would be called an 'Event'

@table @samp
@item @code{Barcheck_req}
    Checks during music processing if start of this voice element
    coincides with the start of a measure. Handy to check if you left out
    some voice elts.
@item @code{Note_req}
    LilyPond has to decide if the ball should be hanging left or
    right. This influences the horizontal dimensions of a column, and this
    is why request processing should be done before horizontal spacing.
    Other voices' frivolities may cause the need for accidentals, so this
    is also for the to decide. The engraver can decide on positioning based on
    ottava commands and the appropriate clef.
@item @code{Rest_req}
    Typeset a rest.
@item @code{Span_req}
    This type of request typically results in the creation of a @code{Spanner}
@item @code{Beam_req}
    Start/stop a beam.
    Engraver has to combine this request with the stem_request, since the
    number of flags that a stem wants to carry will determine the
    number of beams.
@item @code{Dynamic}
    Each dynamic is bound to one note (a crescendo spanning multiple
    notes is thought to be made of two "dynamics": a start and a stop).
    Dynamic changes can occur in a smaller time than the length of its
    note, therefore fore each  @code{Dynamic} request carries a time, measured
    from the start of its note.
@end table

@node Request_engraver, , mudela, Top
@section Request_engraver

In the previous section the idea of Request has been explained, but
this only solves one half of the problem. The other half is deciding
which requests should be honored, which should merged with other
requests, and which should be ignored. Consider this input

@example 

        \type Staff < % chord
                @{ \meter 2/4; [c8 c8] @}
                @{\meter 2/4;  [e8 e8] @}
        >
 
@end example 

Both the cs and es are part of a staff (they are in the same
Voice_group), so they should share meters, but the two [ ] pairs
should be merged.

The judge in this "allocation" problem a set of brokers: the requests
are transmitted to so-called engravers which respond if they want to
accept a request eg, the @code{Notehead_engraver} will accept
@code{Note_req}s, and turn down @code{Slur_req}s. If the Music_iterator
cannot find a engraver that wants the request, it is junked (with a
warning message).

After all requests have been either assigned, or junked, the Engraver
will process the requests (which usually means creating an @code{Item}
or @code{Spanner}). If a @code{Request_engraver} creates something, it
tells the enclosing context. If all items/spanners have been created,
then each Engraver is notified of any created Score_element, via a
broadcasting system.

@unnumberedsubsec example:

@example 

        c4
 
@end example 

produces:

@example 

        Note_request (duration 1/4)
        Stem_request (duration 1/4)
 
@end example 

Note_request will be taken by a @code{Notehead_engraver}, stem_request
will be taken by a @code{Stem_beam_engraver}. @code{Notehead_engraver}
creates a @code{Notehead}, @code{Stem_beam_engraver} creates a
@code{Stem}. Both announce this to the Staff_engraver. Staff_engraver
will tell @code{Stem_beam_engraver} about the @code{Notehead}, which
will add the @code{Notehead} to the @code{Stem} it just created.

To decide on merging, several engravers have been grouped. Please
check @file{init/engraver.ly}.



@node Graphic elements, , , Top 
@section Graphic elements


Music notation is composed of a sets of interrelated glyphs.  In
Lilypond every glyph usually is represented by one object, a so-called
Graphic Object.  The primary relations between graphic objects involve
positions:

@itemize
@item consecutive notes are printed left to right, grouped  in a staff
@item simultaneous notes are horizontally aligned (internally grouped in
a paper column).
@item  the staccato mark is horizontally centered on the note it applies
to.
@end itemize

The abstract encoding of such relations is done with the concept
@dfn{reference point}.  The reference point (in X direction) of the
staccato mark is the note it applies to.  The (X) position of the
staccato mark is stored relative to the position of the note head.  This
means that the staccato will always maintain a fixed offset wrt to the
note head, whereever the head is moved to.

In the same vein, all notes on a staff have their Y positions stored
relative to an abstract object called Axis_group_spanner.  If the
Axis_group_spanner of one staff is moved, the absolute Y positions of
all objects in that spanner change along, in effect causing the staff
and all its contents to move as a whole.

Each graphic object stores a pointer and an relative offset for each
direction: one for the X-axis, one for the Y-axis.  For example, the X
parent of a Note_head usually is a Note_column.  The X parent of a
Note_column usually is either a Collision or a Paper_column. The X
parent of a Collision usually is a Paper_column.  If the Collision
moves, all Note_heads that have that Collision as parent also move, but
the absolute position of the Paper_column does not change.

To build a graphical score with Graphic_elements, conceptually, one
needs to have one Root object (in Lilypond: Line_of_score), and
recursively attach objects to the Root.   However, due to the nature
of the context selection mechanisms, it turns out to be more
advantageous to construct the tree the other way around: first small
trees (note heads, barlines etc.) are created, and these are
subsequently composed into larger trees, which are finally hung on a
Paper_column (in X direction) or Line_of_score (in Y direction). 

The structure of the X,Y parent relations are determined by the
engravers and notation contexts:

The most important X-axis parent relation depends on the timing: notes
that come earlier are attached to Paper_column that will be positioned
more to the left.

The most important Y-axis relation depends on containment of contexts:
notes (created in a Thread or Voice context) are put in the staff where
the originating context lives in.

Graphical_axis_groups are special graphic objects, that are designed to
function as a parent.  The size of a Graphical_axis_groups group is the
union of its children.

@node Score elements, ,  , Top

Besides relative positions there are lots of other relations between
elements. Lilypond does not contain other specialized relation
management (Like the relative positioning code).  In stead, objects can
be connected through dependencies, which sets the order in which objects
are to be processed.

Example: the direction of a beamed stem should equal the direction of
the beam.  When the stem is a added to the beam, a dependency on the
beam is set in the stem: this means that @code{Beam::do_pre_processing
()} (which does various direction related things) will be called before
@code{Stem::do_pre_processing ()}.

The code that manages dependencies resides in the class
@code{Score_element}, a derived class of @code{Graphical_element}.  The
bulk of the code handles line breaking related issues.

To sketch the problems with line breaking: suppose a slur spans a line
break,
@example

c4(  c'''' c | \break d d )d

@end example
In this case, the slur will appear as two parts, the first part spanning
the first three notes (the @code{c}s), the second spanning the last
three (the @code{d}s).  Within Lilypond this is modeled as breaking the
slur in parts: from the Original slur, two new clones of the old slur
are made. Initially both clones depend on the six notes.  After the
hairy code in Score_element, Spanner and Item which does substitutions
in sets of dependencies, the first clone depends on the first three
notes, the second on the last three.

The major derived classes of Score_element are Item and  Spanner.
An item has one horizontal position.  A spanner hangs on two  items.

@node Items, , , Top
@section Items



An item is a score element that is associated with only one
Paper_column. Examples are note heads, clefs, sup and superscripts, etc.
Item is a derived class of Score_element.

The shape of an item is known before the break calculations, and line
spacing depends on the size of items: very wide items need more space
than very small ones.

An additional complication is the behavior of items at linebreaks.  For
example, when you do a time signature change, you get only one symbol.
If it occurs at a linebreak, the new time signature must be printed both
before and after the linebreak.  Other `breakable symbols' such as
clefs, and bar lines also exhibit this behavior. 

if a line of music is broken, the next line usually gets a clef. So in
TeX terms, the clef is a postbreak. The same thing happens with meter
signs: Normally the meter follows the bar. If a line is broken at that
bar, the bar along with the meter stays on the "last" line, but the next
line also gets a meter sign after the clef.   To support this,
breakable items are generated in the three versions: original
(unbroken), left (before line break) and right (after line break).
During the line spacing, these versions are used to try how the spacing
of a  line works out.

Once the definitive spacing is determined, dependencies (and various
other pointers) are substituted such that all dependencies point at the
active items: either they point at the original, or they point at left
and right.

@node Spanners, , , Top
@section Spanners

Spanners are symbols that are of variable shape, eg. Slurs, beams, etc.
Spanners is a derived class of Score_element.

The final shape can only be determined after the line breaking process.
All spanners are spanned on two items, called the left and right
boundary item.  The X reference point is the left boundary item.


@node Future work, , , Top
@section Future work

There are plans to unify Spanner and Item, so there will no longer be
such a clear distinction between the two.  Right now, Score_elements are
always either Item or either Spanner.

Most of the properties of a graphic object are now member variables of
the classes involved.  To offer configurability, we want to move these
variables to scheme (GUILE) variables, and no longer use C++ code to
calculate them, but use Scheme functions.

@node Coding standards,  , , Top

@chapter CodingStyle - standards while programming for GNU
LilyPond

Functions and methods do not return errorcodes.


@unnumberedsubsec Languages

C++ and Python are preferred.  Perl is not.  Python code should use an
indent of 8, using TAB characters.

@unnumberedsubsec Filenames

Definitions of classes that are only accessed via pointers
(*) or references (&) shall not be included as include files.

filenames

@example 
        ".hh"   Include files
        ".cc"   Implementation files
        ".icc"  Inline definition files
        ".tcc"  non inline Template defs
@end example 

in emacs:

@example 
        (setq auto-mode-alist
              (append '(("\\.make$" . makefile-mode)
                        ("\\.cc$" . c++-mode)
                        ("\\.icc$" . c++-mode)
                        ("\\.tcc$" . c++-mode)
                        ("\\.hh$" . c++-mode)
                        ("\\.pod$" . text-mode)         
                        )
                      auto-mode-alist))
@end example 


The class Class_name_abbreviation is coded in @file{class-name-abbr.*}

@unnumberedsubsec Indentation

Standard GNU coding style is used.   In emacs:

@example 
        (add-hook 'c++-mode-hook
                  '(lambda() (c-set-style "gnu")
                     )
                  )
@end example 

If you like using font-lock, you can also add this to your @file{.emacs}:

@example 
        (setq font-lock-maximum-decoration t)
        (setq c++-font-lock-keywords-3 
              (append
               c++-font-lock-keywords-3
               '(("\\b\\([a-zA-Z_]+_\\)\\b" 1 font-lock-variable-name-face)
               ("\\b\\([A-Z]+[a-z_]+\\)\\b" 1 font-lock-type-face))
               ))
@end example 

@unnumberedsubsec Classes and Types

@example 
        This_is_a_class
@end example 

@unnumberedsubsec Members

@example 
        Class::member ()
        Type Class::member_type_
        Type Class::member_type ()
@end example 

the @code{type} is a Hungarian notation postfix for @code{Type}. See below

@unnumberedsubsec Macros

Macros should be written completely in uppercase

The code should not be compilable if proper macro declarations are not
included. 

Don't laugh. It took us a whole evening/night to figure out one of
these bugs, because we had a macro that looked like
@code{DECLARE_VIRTUAL_FUNCTIONS ()}. 

@unnumberedsubsec Broken code

Broken code (hardwired dependencies, hardwired constants, slow
algorithms and obvious limitations) should be marked as such: either
with a verbose TODO, or with a short "ugh" comment.

@unnumberedsubsec Comments

The source is commented in the DOC++ style.  Check out doc++ at
@uref{http://www.zib.de/Visual/software/doc++/index.html}

@example 

        /*
                C style comments for multiline comments.
                They come before the thing to document.
                [...]
        */

        /**
                short description.
                Long class documentation.
                (Hungarian postfix)

                TODO Fix boring_member ()
        */
        class Class @{
                /**
                  short description.
                  long description
                */

                Data data_member_;

                /**
                        short memo. long doco of member ()
                        @@param description of arguments
                        @@return Rettype
                */
                Rettype member (Argtype);

                /// memo only
                boring_member () @{
                        data_member_ = 121; // ugh
                @}
        @};
 
@end example 

        
Unfortunately most of the code isn't really documented that good.

@unnumberedsubsec Members (2)

Standard methods:

@example 

        ///check that *this satisfies its invariants, abort if not.
        void OK () const

        /// print *this (and substructures) to debugging log
        void print () const

        /**
        protected member. Usually invoked by non-virtual XXXX ()
        */
        virtual do_XXXX ()

        /**add some data to *this.
        Presence of these methods usually imply that it is not feasible to this
        via  a constructor
        */
        add (..)

        /// replace some data of *this
        set (..)
 
@end example 

@unnumberedsubsec Constructor

Every class should have a default constructor.  

Don't use non-default constructors if this can be avoided:

@example 

        Foo f(1)
 
@end example 

is less readable than

@example 

        Foo f;
        f.x = 1
 
@end example 

or 

@example 

        Foo f(Foo_convert::int_to_foo (1))
 
@end example 

@unnumberedsec Hungarian notation naming convention

Proposed is a naming convention derived from the so-called
@emph{Hungarian Notation}.  Macros, @code{enum}s and @code{const}s are all
uppercase, with the parts of the names separated by underscores.

The hungarian notation  is to be used when variables are not declared
near usage (mostly in member variables and functions).

@unnumberedsubsec Types

@table @samp
@item @code{byte}
    unsigned char. (The postfix _by is ambiguous)
@item @code{b}
    bool
@item @code{bi}
    bit
@item @code{ch}
    char
@item @code{f}
    float
@item @code{i}
    signed integer
@item @code{str}
    string class
@item @code{sz}
    Zero terminated c string
@item @code{u}
    unsigned integer
@end table

@unnumberedsubsec User defined types

@example 

        /** Slur blah. blah.
        (slur)
        */
        class Slur @{@};
        Slur* slur_p = new Slur;
 
@end example 

@unnumberedsubsec Modifiers

The following types modify the meaning of the prefix. 
These are preceded by the prefixes:

@table @samp
@item @code{a}
    array
@item @code{array}
    user built array.
@item @code{c}
    const. Note that the proper order is @code{Type const}
    i.s.o. @code{const Type}
@item @code{C}
    A const pointer. This would be equivalent to @code{_c_l}, but since any
    "const" pointer has to be a link (you can't delete a const pointer),
    it is superfluous.
@item @code{l}
    temporary pointer to object (link)
@item @code{p}
    pointer to newed object
@item @code{r}
    reference
@end table

@unnumberedsubsec Adjective

Adjectives such as global and static should be spelled out in full.
They come before the noun that they refer to, just as in normal english.

@example 

foo_global_i: a global variable of type int commonly called "foo".
 
@end example 

static class members do not need the static_ prefix in the name (the
Class::var notation usually makes it clear that it is static)

@table @samp
@item @code{loop_i}
    Variable loop: an integer
@item @code{u}
    Temporary variable: an unsigned integer
@item @code{test_ch}
    Variable test: a character
@item @code{first_name_str}
    Variable first_name: a String class object
@item @code{last_name_ch_a}
    Variable last_name: a @code{char} array
@item @code{foo_i_p}
    Variable foo: an @code{Int*} that you must delete
@item @code{bar_i_l}
    Variable bar: an @code{Int*} that you must not delete
@end table

Generally default arguments are taboo, except for nil pointers.

The naming convention can be quite conveniently memorised, by
expressing the type in english, and abbreviating it

@example 

        static Array<int*> foo
 
@end example 

@code{foo} can be described as "the static int-pointer user-array", so you get

@example 

        foo_static_l_arr
 
@end example 


@unnumberedsec Miscellaneous
    
For some tasks, some scripts are supplied, notably creating patches, a
mirror of the website, generating the header to put over cc and hh
files, doing a release.

Use them.

@node Making patches,  , , Top


@unnumberedsec name
    

PATCHES - track and distribute your code changes

This page documents how to distribute your changes to GNU lilypond
    
We would like to have unified context diffs with full pathnames.  A
script automating supplied with Lily.

Distributing a change normally goes like this:

@itemize @bullet
@item make your fix/add your code 
@item Add changes to CHANGES, and add yourself to Documentation/topdocs/AUTHORS.texi
@item generate a patch, 
@item e-mail your patch to one of the mailing lists
    gnu-music-discuss@@gnu.org or bug-gnu-music@@gnu.org
@end itemize

Please do not send entire files, even if the patch is bigger than the
original.  A patch makes it clear what is changed, and it won't
overwrite previous (not yet released) changes.

@unnumberedsec Generating a patch

Simple version: run

@example
        make -C lilypond-x.y.z/ distclean
        make -C lilypond-x.y.z.NEW/ distclean
        diff -urN lilypond-x.y.z/ lilypond-x.y.z.NEW/
@end example

Complicated (but automated) version:

In @file{VERSION}, set MY_PATCH_LEVEL:

@example 

    VERSION:
        ...
        MY_PATCH_LEVEL=jcn1
 
@end example 

In @file{CHANGES}, enter a summary of changes:

@example 
        pl 0.1.73.jcn1
                - added PATCHES.texi
@end example 

Then, from the top of Lily's source tree, type

@example 
    make release
@end example 

These handy python scripts assume a directory structure which looks
like:

@example 

    lilypond -> lilypond-x.y.z   # symlink to development directory
    lilypond-x.y.z/              # current development
    patches/                     # patches between different releases
    releases/                    # .tar.gz releases

@end example 

(Some scripts also assume this lives in  @file{$HOME/usr/src}).

        
@unnumberedsec Applying patches
    

If you're following LilyPond development regularly, you probably want to
download just the patch for each subsequent release.
After downloading the patch (into the patches directory, of course), simply 
apply it:

@example 

    gzip -dc ../patches/lilypond-0.1.74.diff.gz | patch -p1 -E
 
@end example 

and don't forget to make automatically generated files:

@example 

    autoconf footnote(patches don't include automatically generated files, 
    i.e. file(configure) and files generated by file(configure).)

    configure
 
@end example 
    

@bye