Update at Fri Oct 12 11:00:31 EDT 2018 by tim

[xcircuit.git] / README.notes

XCircuit v3.10
-----------------------------------------------------------------------------

An X11 drawing program [especially for circuit schematics]
Copyright (c) 2017 Tim Edwards
August 1993 -- April 2017
-----------------------------------------------------------------------------

This README file contains additional notes about the program which may or may
not be of interest to anyone.
-----------------------------------------------------------------------------

ASG:

   Versions beginning with 3.2 can be compiled with the ASG package,
or "Automatic Schematic Generation".  The package is derived from SPAR,
written by Stephen Frezza (currently at Gannon University).  SPAR is
based on the same "corner-stitched tile" structures used by the EDA
tool "magic", modified to encompass the probabilistic nature of auto-
routing.  The ASG package currently supports (to a limited extent)
reading of SPICE files, but will eventually support EDIF netlist output
from Verilog (and possibly VHDL) for turning netlists into schematics.
Also, the ASG package will handle constrained moving, re-routing networks
in response to moving elements around the schematic drawing.

Toolscript:

   Versions beginning with 3.0 are part of a methodology known as
"ScriptEDA", a call to put major open-source EDA tools into a similar
framework under an interpreter, where most upper-level functions can be
written as scripts.  My more narrowly-defined version of ScriptEDA is
called "toolscript".  The rationale behind "toolscript" is to allow a
single interpreter to launch multiple EDA tools, such that a
single script can control multiple programs, bringing together layout
tools, schematic capture, simulation, and so forth, in a highly
integrated manner.  "Toolscript" requires that each EDA tool should be
a shared-object library of routines, that each tool should have its
own command namespace, that each tool should launch itself from a
"start" command in its namespace, and that the "start" command
should return control to the interpreter such that the interpreter is
always available from the calling terminal.  Applications are encouraged
to rely heavily on scripts for functions which do not require fast
execution.  In particular, scripts should handle most input/output
functions, so that users can easily change file formats to suit their
needs.  The GUI for toolscript applications should be written in Tk,
which integrates the graphics event loop with the interpreter, greatly
simplifying handling of the interaction between the two.  To synchronize
the GUI with commands called from the command-line, each program should
have a "command tag" function which can associate a script procedure
with each command.  These callback procedures can then be defined within
the GUI script to update the GUI in response to commands.  Finally,
each application should redefine "printf" output statements to call the
corresponding output method for the intperpreter (e.g., "Tcl_Eval" for
compatibility with the "tkcon" console).  This allows error and standard
output to be handled independently.  Within "tkcon", stdout and stderr
are printed in blue and red, respectively, making the program output
easier to read.

   The "toolscript" concept is ideally interpreter-independent.  However,
due mostly to ease of implementation and syntactical issues, TCL has
ended up as the only supported interpreter.  TCL calls are heavily
embedded in the source code, making it somewhat difficult to simply
add another interpreter.  While XCircuit can be compiled with the Python
interpreter, it is embedded, not extended, and so does not follow the
principles of "toolscript".

Rationale:

   This program was written with a number of things in mind.  The major
point in writing it was that I found all of the drawing programs freely
distributed at the usual X11 ftp sources (ftp.uu.net, export.lcs.mit.edu,
etc.), which includes xpic and xfig, and even some other programs like
idraw and commercial programs like framemaker, were lacking in some manner.
Some of them, like xpic, are missing fundamental parts, while others, like
xfig and framemaker, have a cumbersome user interface and require
inordinately large amounts of time to create simple pictures.  The really
powerful programs like Adobe Illustrator are, naturally, expensive.
There has been a long-standing tradition of making MacDraw lookalikes which
I find particularly disturbing, since a number of features of that GUI are
fundamentally screwed up, including the dependence on a black-and-white color
scheme for things such as grid lines and edit points which can make the screen
cluttered and difficult to work with.  I have chosen to freely ignore all such
standards in cases where they seem to hinder rather than help the user.
   Disclaimers:  The above comments refer to versions of the abovementioned
programs available as of 1993 and do not necessarily reflect the programs'
current status.  Particularly the latest (from 1995 or so) version of xfig 
is quite good, though it continues to have a "MacDraw" feel to it.  XCircuit's
schematic capture has always been secondary to its use for drawing, and
users may prefer programs such as "gschem" (by Ales Hvezda, part of gEDA)
that are more in keeping with "normal" standards for EDA software.  Because
such programs exist, XCircuit's schematic capture, like its basic drawing
functions, often deviate from the norm to explore radical ideas.  Among
XCircuit's "radical ideas" are automatic port finding, its parameterization
method, embedded output formatting directives, and (a recent addition)
automatic schematic (re)generation.

The principles of XCircuit are as follows:

A) Structural backbone
   1) There are six drawing elements which are considered "primary".  
      These are as follows:  

      a) polygon (multiple lines which may or may not be closed and filled)
      b) arc (ellipse segment which may be closed and/or filled as above)
      c) label (any text)
      d) curve (based on the PostScript "curveto" algorithm)
      e) graphic (graphic image, e.g., from a PPM file)
      f) object instance (see below)

   2) There are two drawing elements which are considered "secondary", which
      is to say, contain instances of primary elements.  These are:

      a) path (a connected series of polygons, arcs, and/or curves)
      b) object (something containing polygons, arcs, labels, 
                 curves, paths, and instances of other objects)

      Each object is instanced by a primary element called the "object
      instance", a self-contained unit which allows an object to be scaled, 
      rotated, and translated within another object.  Since objects contain
      object instances and object instances point to objects, there is a
      two-step recursion which is central to the concept of xcircuit.

   3) There is one primary object instance called the TopObject, which is
      the program window, and its associated object, which is whatever
      appears on the screen.  TopObject is a pointer, so that any object
      in memory may be the top object, allowing a push/pop edit stack.
      All other objects are organized in an array of pointers called the
      Library.  The Library is filled on startup with the contents of a
      Library file, and is expanded whenever a new object is created.
      
      Actually, there are many different TopObjects, called "pages",
      which can be switched between at will by typing the number of
      the page, 1 through 9 and 0 for page 10, or through the "Page"
      menu command, which allows access to as many pages as are
      defined.  The "load" and "save" commands act only on the
      present page (unless multiple pages are loaded at a time), as
      do all the drawing commands.

   4) All elements allocate memory as they are created and release the
      memory as they are destroyed, so the only limit on the number of
      elements in a drawing is the memory of the workstation.

B) Input/Output

   1) All input and output is in PostScript, albeit a version of PostScript
      with extremely restricted syntax:  xcircuit is in no sense a PS
      interpreter!  The primary element types are defined as procedures
      in the PostScript prologue;  all secondary elements are defined as
      "def" structures in PostScript, the only restriction being that no
      definition can be called before it is defined.

   2) Output can be encapsulated or non-encapsulated ("full page"), and
      can have multiple pages.  All pages which declare the same
      "filename" on the "output properties" window will be saved to
      the same file.

   3) The current state of the program is captured from time to time
      (default 10 minutes or 20 "significant" changes to the drawing)
      in a temporary file.  If XCircuit crashes, work can be recovered
      by running "xcircuit" without arguments (i.e., without a specific
      filename on the command line).  This method makes the filename
      transparent to the user, but the file can be found (on most
      systems) as /tmp/XC??? where ??? is a randomly-generated number
      that uniquely identifies the file.

   4) Text is vector-drawn, in order to best match the flexibility of
      PostScript, which unlike X11 allows text to be scaled to any size
      and rotated to any orientation (modern font renderers can do this,
      but are not good at fast on-the-fly rendering needed for a drawing
      The quality of the vector-drawn text on the screen is relatively
      poor;  however, the end product of xcircuit is the PostScript file,
      which does not suffer from those limitations.  Text vectors are
      xcircuit elements read from an xcircuit library file, with
      encoding information read from a separate file.

   5) X11 does not have any capability for generalized curves (splines).
      Therefore, splines are constructed by an N-segment approximation
      which internally looks like a polygon.  The PostScript output is a 
      smooth curve instead of the approximation.

   6) X11 does not have any capability for generalized ellipses (ellipses
      under any rotation).  Similarly to the splines, arcs are constructed
      by an N-segment approximation.

   7) Implementing all arcs and curves as N-segment approximation allows
      the possibility of general paths, or sequences of polygons, arc
      sections, and curves which form a path that can be bordered, closed
      and/or filled like an individual arc, curve, or polygon.

   8) Other than the qualitative appearance of the text, the screen output
      of xcircuit is meant to be the closest possible match to the
      rendering of the PostScript output.  A PostScript hack allows
      stipple pattern fills to be effectively "transparent" (something
      that PostScript does not do naturally).  However, the effect is
      *extremely* device-dependent because it overrides the printer's
      halftoning operators.  One effect of this is that halftoning
      will look abnormally large when viewed at screen resolutions,
      e.g., with GhostScript.

C) User Interface

   1) The mouse button system and the object library are based on the
      Caltech circuit-simulation program "log" (either "analog" or "diglog").
      As is usual for most applications, mouse button 1 is used to select
      things or cycle between selections, button 2 usually terminates a
      command, and button 3 will abort a command.  In addition, button 1
      has a function called "hold", which occurs after the button has
      been continually pressed for a short length of time (about 1/5 second).
      As the name implies, "hold" grabs hold of an element which can then
      be moved around the screen.

      Because this program is tailored to circuit drawing, the most common
      functions are drawing lines and moving object instances.  Therefore,
      these two functions are made directly available from the mouse
      button 1 (after a "tap" or a "hold", respectively).  The next most
      common function is selection of elements singly or in groups.
      Button 2 can be used to select an item if tapped, and if pressed
      and held down, a box will be drawn and everything inside that box
      selected when the button is released.  All other commands are 
      available from the pulldown menus and/or from the keyboard as
      single-key macros.  Keyboard macros are quicker, since they act
      on the present cursor position, whereas menu commands require an
      extra step.

      From version 3.4, buttons and keys are handled identically by the
      interface, with the "hold" function becoming a modifier like "shift"
      or "control".   All XCircuit functions can be arbitrarily bound to
      any button or key.  The button 1 binding corresponds to the drawing
      mode.  If button 1 is bound to function "Wire", then it works as
      described above.  If it is bound to "Arc", then button 1 can be
      used to draw arcs, while the other functions remain the same.

   2) The libraries are intended to provide a convenient way to store and
      retrieve elements of a picture which will be used more than once.
      For the application of circuit drawing, a built-in library provides
      basic objects such as transistors, amplifiers, resistors, capacitors,
      arrows, circles, power and ground symbols, and the like.  This file
      ("builtins.lps") is loaded on startup;  the program first looks in
      the current directory, and then searches in a given global directory
      (something like ~cad/lib/xcircuit/) if it could not find it initially.
      Thus each user can add to or modify the file of builtins to reflect
      personal taste.  Since the PostScript output contains all object
      definitions, these changes to the built-in functions are inherently
      transferrable.  The program checks objects with the same name for
      equivalence of their contents, and changes names as necessary to
      avoid conflicts.

      An additional library, the user library, contains all of the
      objects which are created (using the "make object" command)
      during a session or objects which are loaded from a file but
      do not belong to the built-in library.  The user library can
      be saved by itself and reloaded later either into the user
      library or into the built-in library.  This is the way to
      create a personalized built-in library set.

   3) In the non-TCL-based version of XCircuit, the special file
      "startup.script" can be edited to include lines such as the
      following:

                filename loadlibrary

      The loadlibrary command will automatically include another
      library file.  This method makes it convenient to divide up
      libraries into separate files according to (for instance)
      function (digital, analog, clip art, etc.), and use the
      loadlibrary commands to choose which ones will be loaded
      on startup.  Caveat:  If the order of loading is such that
      an object contains an instance of another object which has
      not yet been loaded, xcircuit will complain.

      In the TCL-based version of XCircuit, the equivalent file is
      "xcstartup.tcl" and the syntax is:

        library n load filename

      where "filename" is the name of the library file (e.g.,
      "digital.lps"), and "n" is the library page number.  By
      specifying the same page as an existing library, a set of
      objects can be appended to that library.  "n" may also refer
      to a library by name.

      By default, libraries have the extension ".lps".  As they
      contain only macros, they cannot be displayed with any
      PostScript interpreter.

   4) Xcircuit has a strict sense of zooming and snap-to grid.  Objects
      scale absolutely;  line widths and text sizes will increase/decrease
      proportionally with the zoom, as well as dot/dash spacing and all
      other features.  Xcircuit does have a minimum integer grid in
      coordinate space, which translates to 0.005 inches at an output scale
      of 1.  The maximum zoom scale gives a screen size translating to
      about 100 by 100 inches at an output scale of 1.  The effective
      scale can be varied by changing the output scale (reached from the
      "File/Write" menu selection) in order to fit a drawing to a page or
      to get a grid matched to a specific dimension.  A separate scale
      parameter changes the scale of the reported position relative to
      the output scale (as it will appear on a printed page).

   5) The snap-to grid is an all-important feature for circuit drawing,
      wherein it is critical that wires line up properly with one another.
      In xcircuit, there is no way to get off the snap-to grid except by
      turning the snap function off and physically pushing elements off
      the grid.  In any case, objects can always be returned to the snap
      grid with the "snap" function (key 'S').

D) Schematic Capture

      XCircuit implements a sophisticated schematic capture.  Unlike
      virtually all schematic capture software currently available,
      xcircuit allows the designer to draw the circuit in a "natural"
      way, making use both of schematic hierarchies and simple
      hierarchies created simply by grouping elements together.  If
      the chosen netlist format is hierarchical (like SPICE), both
      hierarchical forms will be retained in the output.  For element
      grouping, input/output ports connecting into to the group will
      be determined automatically, from context.

      Individual networks can be queried (key 'Alt-W') to highlight
      the entire visible part of the network, with the network name
      printed in the message window. 

      Any schematic can cover multiple pages.  One page is marked as
      "primary" and the others marked as "secondary".  Netlists are
      generated in the primary page object only, but incorporate
      information from all the schematic pages.  All pages which have
      the same "page label" are considered to be part of the same
      schematic.

-----------------------------------------------------------------------------

Xcircuit reference:
   Local:  See the xcircuit manpage ("man xcircuit")
   Online: http://opencircuitdesign.com/xcircuit/features.html

-----------------------------------------------------------------------------

Online tutorials: See
   Standard:          http://opencircuitdesign.com/xcircuit/tutorial/tutorial.html
   Schematic capture: http://opencircuitdesign.com/xcircuit/tutorial/tutorial2.html

-----------------------------------------------------------------------------

Things to do: See the file "TODO"
          
-----------------------------------------------------------------------------