xref: /freebsd/contrib/ncurses/doc/html/hackguide.html (revision ae83180158c4c937f170e31eff311b18c0286a93)
1<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 3.0//EN">
2<!--
3  $Id: hackguide.html,v 1.25 2000/03/25 18:45:21 tom Exp $
4-->
5<HTML>
6<HEAD>
7<TITLE>A Hacker's Guide to Ncurses Internals</TITLE>
8<link rev="made" href="mailto:bugs-ncurses@gnu.org">
9<!--
10This document is self-contained, *except* that there is one relative link to
11the ncurses-intro.html document, expected to be in the same directory with
12this one.
13-->
14</HEAD>
15<BODY>
16
17<H1>A Hacker's Guide to NCURSES</H1>
18
19<H1>Contents</H1>
20<UL>
21<LI><A HREF="#abstract">Abstract</A>
22<LI><A HREF="#objective">Objective of the Package</A>
23<UL>
24<LI><A HREF="#whysvr4">Why System V Curses?</A>
25<LI><A HREF="#extensions">How to Design Extensions</A>
26</UL>
27<LI><A HREF="#portability">Portability and Configuration</A>
28<LI><A HREF="#documentation">Documentation Conventions</A>
29<LI><A HREF="#bugtrack">How to Report Bugs</A>
30<LI><A HREF="#ncurslib">A Tour of the Ncurses Library</A>
31<UL>
32<LI><A HREF="#loverview">Library Overview</A>
33<LI><A HREF="#engine">The Engine Room</A>
34<LI><A HREF="#input">Keyboard Input</A>
35<LI><A HREF="#mouse">Mouse Events</A>
36<LI><A HREF="#output">Output and Screen Updating</A>
37</UL>
38<LI><A HREF="#fmnote">The Forms and Menu Libraries</A>
39<LI><A HREF="#tic">A Tour of the Terminfo Compiler</A>
40<UL>
41<LI><A HREF="#nonuse">Translation of Non-<STRONG>use</STRONG> Capabilities</A>
42<LI><A HREF="#uses">Use Capability Resolution</A>
43<LI><A HREF="#translation">Source-Form Translation</A>
44</UL>
45<LI><A HREF="#utils">Other Utilities</A>
46<LI><A HREF="#style">Style Tips for Developers</A>
47<LI><A HREF="#port">Porting Hints</A>
48</UL>
49
50<H1><A NAME="abstract">Abstract</A></H1>
51
52This document is a hacker's tour of the <STRONG>ncurses</STRONG> library and utilities.
53It discusses design philosophy, implementation methods, and the
54conventions used for coding and documentation.  It is recommended
55reading for anyone who is interested in porting, extending or improving the
56package.
57
58<H1><A NAME="objective">Objective of the Package</A></H1>
59
60The objective of the <STRONG>ncurses</STRONG> package is to provide a free software API for
61character-cell terminals and terminal emulators with the following
62characteristics:
63
64<UL>
65<LI>Source-compatible with historical curses implementations (including
66     the original BSD curses and System V curses.
67<LI>Conformant with the XSI Curses standard issued as part of XPG4 by
68     X/Open.
69<LI>High-quality -- stable and reliable code, wide portability, good
70     packaging, superior documentation.
71<LI>Featureful -- should eliminate as much of the drudgery of C interface
72     programming as possible, freeing programmers to think at a higher
73     level of design.
74</UL>
75
76These objectives are in priority order.  So, for example, source
77compatibility with older version must trump featurefulness -- we cannot
78add features if it means breaking the portion of the API corresponding
79to historical curses versions.
80
81<H2><A NAME="whysvr4">Why System V Curses?</A></H2>
82
83We used System V curses as a model, reverse-engineering their API, in
84order to fulfill the first two objectives. <P>
85
86System V curses implementations can support BSD curses programs with
87just a recompilation, so by capturing the System V API we also
88capture BSD's. <P>
89
90More importantly for the future, the XSI Curses standard issued by X/Open
91is explicitly and closely modeled on System V.  So conformance with
92System V took us most of the way to base-level XSI conformance.
93
94<H2><A NAME="extensions">How to Design Extensions</A></H2>
95
96The third objective (standards conformance) requires that it be easy to
97condition source code using <STRONG>ncurses</STRONG> so that the absence of nonstandard
98extensions does not break the code. <P>
99
100Accordingly, we have a policy of associating with each nonstandard extension
101a feature macro, so that ncurses client code can use this macro to condition
102in or out the code that requires the <STRONG>ncurses</STRONG> extension. <P>
103
104For example, there is a macro <CODE>NCURSES_MOUSE_VERSION</CODE> which XSI Curses
105does not define, but which is defined in the <STRONG>ncurses</STRONG> library header.
106You can use this to condition the calls to the mouse API calls.
107
108<H1><A NAME="portability">Portability and Configuration</A></H1>
109
110Code written for <STRONG>ncurses</STRONG> may assume an ANSI-standard C compiler and
111POSIX-compatible OS interface.  It may also assume the presence of a
112System-V-compatible <EM>select(2)</EM> call. <P>
113
114We encourage (but do not require) developers to make the code friendly
115to less-capable UNIX environments wherever possible. <P>
116
117We encourage developers to support OS-specific optimizations and methods
118not available under POSIX/ANSI, provided only that:
119
120<UL>
121<LI>All such code is properly conditioned so the build process does not
122     attempt to compile it under a plain ANSI/POSIX environment.
123<LI>Adding such implementation methods does not introduce incompatibilities
124     in the <STRONG>ncurses</STRONG> API between platforms.
125</UL>
126
127We use GNU <CODE>autoconf(1)</CODE> as a tool to deal with portability issues.
128The right way to leverage an OS-specific feature is to modify the autoconf
129specification files (configure.in and aclocal.m4) to set up a new feature
130macro, which you then use to condition your code.
131
132<H1><A NAME="documentation">Documentation Conventions</A></H1>
133
134There are three kinds of documentation associated with this package.  Each
135has a different preferred format:
136
137<UL>
138<LI>Package-internal files (README, INSTALL, TO-DO etc.)
139<LI>Manual pages.
140<LI>Everything else (i.e., narrative documentation).
141</UL>
142
143Our conventions are simple:
144<OL>
145<LI><STRONG>Maintain package-internal files in plain text.</STRONG>
146     The expected viewer for them <EM>more(1)</EM> or an editor window; there's
147     no point in elaborate mark-up.
148
149<LI><STRONG>Mark up manual pages in the man macros.</STRONG>  These have to be viewable
150     through traditional <EM>man(1)</EM> programs.
151
152<LI><STRONG>Write everything else in HTML.</STRONG>
153</OL>
154
155When in doubt, HTMLize a master and use <EM>lynx(1)</EM> to generate
156plain ASCII (as we do for the announcement document). <P>
157
158The reason for choosing HTML is that it's (a) well-adapted for on-line
159browsing through viewers that are everywhere; (b) more easily readable
160as plain text than most other mark-ups, if you don't have a viewer; and (c)
161carries enough information that you can generate a nice-looking printed
162version from it.  Also, of course, it make exporting things like the
163announcement document to WWW pretty trivial.
164
165<H1><A NAME="bugtrack">How to Report Bugs</A></H1>
166
167The <A NAME="bugreport">reporting address for bugs</A> is
168<A HREF="mailto:bug-ncurses@gnu.org">bug-ncurses@gnu.org</A>.
169This is a majordomo list; to join, write
170to <CODE>bug-ncurses-request@gnu.org</CODE> with a message containing the line:
171<PRE>
172             subscribe &lt;name&gt;@&lt;host.domain&gt;
173</PRE>
174
175The <CODE>ncurses</CODE> code is maintained by a small group of
176volunteers.  While we try our best to fix bugs promptly, we simply
177don't have a lot of hours to spend on elementary hand-holding.  We rely
178on intelligent cooperation from our users.  If you think you have
179found a bug in <CODE>ncurses</CODE>, there are some steps you can take
180before contacting us that will help get the bug fixed quickly. <P>
181
182In order to use our bug-fixing time efficiently, we put people who
183show us they've taken these steps at the head of our queue.  This
184means that if you don't, you'll probably end up at the tail end and
185have to wait a while.
186
187<OL>
188<LI>Develop a recipe to reproduce the bug.
189<p>
190Bugs we can reproduce are likely to be fixed very quickly, often
191within days.  The most effective single thing you can do to get a
192quick fix is develop a way we can duplicate the bad behavior --
193ideally, by giving us source for a small, portable test program that
194breaks the library. (Even better is a keystroke recipe using one of
195the test programs provided with the distribution.)
196
197<LI>Try to reproduce the bug on a different terminal type. <P>
198
199In our experience, most of the behaviors people report as library bugs
200are actually due to subtle problems in terminal descriptions.  This is
201especially likely to be true if you're using a traditional
202asynchronous terminal or PC-based terminal emulator, rather than xterm
203or a UNIX console entry. <P>
204
205It's therefore extremely helpful if you can tell us whether or not your
206problem reproduces on other terminal types.  Usually you'll have both
207a console type and xterm available; please tell us whether or not your
208bug reproduces on both. <P>
209
210If you have xterm available, it is also good to collect xterm reports for
211different window sizes.  This is especially true if you normally use an
212unusual xterm window size -- a surprising number of the bugs we've seen
213are either triggered or masked by these.
214
215<LI>Generate and examine a trace file for the broken behavior. <P>
216
217Recompile your program with the debugging versions of the libraries.
218Insert a <CODE>trace()</CODE> call with the argument set to <CODE>TRACE_UPDATE</CODE>.
219(See <A HREF="ncurses-intro.html#debugging">"Writing Programs with
220NCURSES"</A> for details on trace levels.)
221Reproduce your bug, then look at the trace file to see what the library
222was actually doing. <P>
223
224Another frequent cause of apparent bugs is application coding errors
225that cause the wrong things to be put on the virtual screen.  Looking
226at the virtual-screen dumps in the trace file will tell you immediately if
227this is happening, and save you from the possible embarrassment of being
228told that the bug is in your code and is your problem rather than ours. <P>
229
230If the virtual-screen dumps look correct but the bug persists, it's
231possible to crank up the trace level to give more and more information
232about the library's update actions and the control sequences it issues
233to perform them.  The test directory of the distribution contains a
234tool for digesting these logs to make them less tedious to wade
235through. <P>
236
237Often you'll find terminfo problems at this stage by noticing that the
238escape sequences put out for various capabilities are wrong.  If not,
239you're likely to learn enough to be able to characterize any bug in
240the screen-update logic quite exactly.
241
242<LI>Report details and symptoms, not just interpretations. <P>
243
244If you do the preceding two steps, it is very likely that you'll discover
245the nature of the problem yourself and be able to send us a fix.  This
246will create happy feelings all around and earn you good karma for the first
247time you run into a bug you really can't characterize and fix yourself. <P>
248
249If you're still stuck, at least you'll know what to tell us.  Remember, we
250need details.  If you guess about what is safe to leave out, you are too
251likely to be wrong. <P>
252
253If your bug produces a bad update, include a trace file.  Try to make
254the trace at the <EM>least</EM> voluminous level that pins down the
255bug.  Logs that have been through tracemunch are OK, it doesn't throw
256away any information (actually they're better than un-munched ones because
257they're easier to read). <P>
258
259If your bug produces a core-dump, please include a symbolic stack trace
260generated by gdb(1) or your local equivalent. <P>
261
262Tell us about every terminal on which you've reproduced the bug -- and
263every terminal on which you can't.  Ideally, sent us terminfo sources
264for all of these (yours might differ from ours). <P>
265
266Include your ncurses version and your OS/machine type, of course!  You can
267find your ncurses version in the <CODE>curses.h</CODE> file.
268</OL>
269
270If your problem smells like a logic error or in cursor movement or
271scrolling or a bad capability, there are a couple of tiny test frames
272for the library algorithms in the progs directory that may help you
273isolate it.  These are not part of the normal build, but do have their
274own make productions.  <P>
275
276The most important of these is <CODE>mvcur</CODE>, a test frame for the
277cursor-movement optimization code.  With this program, you can see
278directly what control sequences will be emitted for any given cursor
279movement or scroll/insert/delete operations.  If you think you've got
280a bad capability identified, you can disable it and test again. The
281program is command-driven and has on-line help. <P>
282
283If you think the vertical-scroll optimization is broken, or just want to
284understand how it works better, build <CODE>hashmap</CODE> and read the
285header comments of <CODE>hardscroll.c</CODE> and <CODE>hashmap.c</CODE>; then try
286it out. You can also test the hardware-scrolling optimization separately
287with <CODE>hardscroll</CODE>. <P>
288
289There's one other interactive tester, <CODE>tctest</CODE>, that exercises
290translation between termcap and terminfo formats.  If you have a serious
291need to run this, you probably belong on our development team!
292
293<H1><A NAME="ncurslib">A Tour of the Ncurses Library</A></H1>
294
295<H2><A NAME="loverview">Library Overview</A></H2>
296
297Most of the library is superstructure -- fairly trivial convenience
298interfaces to a small set of basic functions and data structures used
299to manipulate the virtual screen (in particular, none of this code
300does any I/O except through calls to more fundamental modules
301described below).  The files
302<blockquote>
303<CODE>
304lib_addch.c
305lib_bkgd.c
306lib_box.c
307lib_chgat.c
308lib_clear.c
309lib_clearok.c
310lib_clrbot.c
311lib_clreol.c
312lib_colorset.c
313lib_data.c
314lib_delch.c
315lib_delwin.c
316lib_echo.c
317lib_erase.c
318lib_gen.c
319lib_getstr.c
320lib_hline.c
321lib_immedok.c
322lib_inchstr.c
323lib_insch.c
324lib_insdel.c
325lib_insstr.c
326lib_instr.c
327lib_isendwin.c
328lib_keyname.c
329lib_leaveok.c
330lib_move.c
331lib_mvwin.c
332lib_overlay.c
333lib_pad.c
334lib_printw.c
335lib_redrawln.c
336lib_scanw.c
337lib_screen.c
338lib_scroll.c
339lib_scrollok.c
340lib_scrreg.c
341lib_set_term.c
342lib_slk.c
343lib_slkatr_set.c
344lib_slkatrof.c
345lib_slkatron.c
346lib_slkatrset.c
347lib_slkattr.c
348lib_slkclear.c
349lib_slkcolor.c
350lib_slkinit.c
351lib_slklab.c
352lib_slkrefr.c
353lib_slkset.c
354lib_slktouch.c
355lib_touch.c
356lib_unctrl.c
357lib_vline.c
358lib_wattroff.c
359lib_wattron.c
360lib_window.c
361</CODE>
362</blockquote>
363are all in this category.  They are very
364unlikely to need change, barring bugs or some fundamental
365reorganization in the underlying data structures. <P>
366
367These files are used only for debugging support:
368<blockquote>
369<code>
370lib_trace.c
371lib_traceatr.c
372lib_tracebits.c
373lib_tracechr.c
374lib_tracedmp.c
375lib_tracemse.c
376trace_buf.c
377</code>
378</blockquote>
379It is rather unlikely you will ever need to change these, unless
380you want to introduce a new debug trace level for some reasoon.<P>
381
382There is another group of files that do direct I/O via <EM>tputs()</EM>,
383computations on the terminal capabilities, or queries to the OS
384environment, but nevertheless have only fairly low complexity.  These
385include:
386<blockquote>
387<code>
388lib_acs.c
389lib_beep.c
390lib_color.c
391lib_endwin.c
392lib_initscr.c
393lib_longname.c
394lib_newterm.c
395lib_options.c
396lib_termcap.c
397lib_ti.c
398lib_tparm.c
399lib_tputs.c
400lib_vidattr.c
401read_entry.c.
402</code>
403</blockquote>
404They are likely to need revision only if
405ncurses is being ported to an environment without an underlying
406terminfo capability representation. <P>
407
408These files
409have serious hooks into
410the tty driver and signal facilities:
411<blockquote>
412<code>
413lib_kernel.c
414lib_baudrate.c
415lib_raw.c
416lib_tstp.c
417lib_twait.c
418</code>
419</blockquote>
420If you run into porting snafus
421moving the package to another UNIX, the problem is likely to be in one
422of these files.
423The file <CODE>lib_print.c</CODE> uses sleep(2) and also
424falls in this category.<P>
425
426Almost all of the real work is done in the files
427<blockquote>
428<code>
429hardscroll.c
430hashmap.c
431lib_addch.c
432lib_doupdate.c
433lib_getch.c
434lib_mouse.c
435lib_mvcur.c
436lib_refresh.c
437lib_setup.c
438lib_vidattr.c
439</code>
440</blockquote>
441Most of the algorithmic complexity in the
442library lives in these files.
443If there is a real bug in <STRONG>ncurses</STRONG> itself, it's probably here.
444We'll tour some of these files in detail
445below (see <A HREF="#engine">The Engine Room</A>). <P>
446
447Finally, there is a group of files that is actually most of the
448terminfo compiler.  The reason this code lives in the <STRONG>ncurses</STRONG>
449library is to support fallback to /etc/termcap.  These files include
450<blockquote>
451<code>
452alloc_entry.c
453captoinfo.c
454comp_captab.c
455comp_error.c
456comp_hash.c
457comp_parse.c
458comp_scan.c
459parse_entry.c
460read_termcap.c
461write_entry.c
462</code>
463</blockquote>
464We'll discuss these in the compiler tour.
465
466<H2><A NAME="engine">The Engine Room</A></H2>
467
468<H3><A NAME="input">Keyboard Input</A></H3>
469
470All <CODE>ncurses</CODE> input funnels through the function
471<CODE>wgetch()</CODE>, defined in <CODE>lib_getch.c</CODE>.  This function is
472tricky; it has to poll for keyboard and mouse events and do a running
473match of incoming input against the set of defined special keys. <P>
474
475The central data structure in this module is a FIFO queue, used to
476match multiple-character input sequences against special-key
477capabilities; also to implement pushback via <CODE>ungetch()</CODE>. <P>
478
479The <CODE>wgetch()</CODE> code distinguishes between function key
480sequences and the same sequences typed manually by doing a timed wait
481after each input character that could lead a function key sequence.
482If the entire sequence takes less than 1 second, it is assumed to have
483been generated by a function key press. <P>
484
485Hackers bruised by previous encounters with variant <CODE>select(2)</CODE>
486calls may find the code in <CODE>lib_twait.c</CODE> interesting.  It deals
487with the problem that some BSD selects don't return a reliable
488time-left value.  The function <CODE>timed_wait()</CODE> effectively
489simulates a System V select.
490
491<H3><A NAME="mouse">Mouse Events</A></H3>
492
493If the mouse interface is active, <CODE>wgetch()</CODE> polls for mouse
494events each call, before it goes to the keyboard for input.  It is
495up to <CODE>lib_mouse.c</CODE> how the polling is accomplished; it may vary
496for different devices. <P>
497
498Under xterm, however, mouse event notifications come in via the keyboard
499input stream.  They are recognized by having the <STRONG>kmous</STRONG> capability
500as a prefix.  This is kind of klugey, but trying to wire in recognition of
501a mouse key prefix without going through the function-key machinery would
502be just too painful, and this turns out to imply having the prefix somewhere
503in the function-key capabilities at terminal-type initialization. <P>
504
505This kluge only works because <STRONG>kmous</STRONG> isn't actually used by any
506historic terminal type or curses implementation we know of.  Best
507guess is it's a relic of some forgotten experiment in-house at Bell
508Labs that didn't leave any traces in the publicly-distributed System V
509terminfo files.  If System V or XPG4 ever gets serious about using it
510again, this kluge may have to change. <P>
511
512Here are some more details about mouse event handling: <P>
513
514The <CODE>lib_mouse()</CODE>code is logically split into a lower level that
515accepts event reports in a device-dependent format and an upper level that
516parses mouse gestures and filters events.  The mediating data structure is a
517circular queue of event structures. <P>
518
519Functionally, the lower level's job is to pick up primitive events and
520put them on the circular queue.  This can happen in one of two ways:
521either (a) <CODE>_nc_mouse_event()</CODE> detects a series of incoming
522mouse reports and queues them, or (b) code in <CODE>lib_getch.c</CODE> detects the
523<STRONG>kmous</STRONG> prefix in the keyboard input stream and calls _nc_mouse_inline
524to queue up a series of adjacent mouse reports. <P>
525
526In either case, <CODE>_nc_mouse_parse()</CODE> should be called after the
527series is accepted to parse the digested mouse reports (low-level
528events) into a gesture (a high-level or composite event).
529
530<H3><A NAME="output">Output and Screen Updating</A></H3>
531
532With the single exception of character echoes during a <CODE>wgetnstr()</CODE>
533call (which simulates cooked-mode line editing in an ncurses window),
534the library normally does all its output at refresh time. <P>
535
536The main job is to go from the current state of the screen (as represented
537in the <CODE>curscr</CODE> window structure) to the desired new state (as
538represented in the <CODE>newscr</CODE> window structure), while doing as
539little I/O as possible. <P>
540
541The brains of this operation are the modules <CODE>hashmap.c</CODE>,
542<CODE>hardscroll.c</CODE> and <CODE>lib_doupdate.c</CODE>; the latter two use
543<CODE>lib_mvcur.c</CODE>.  Essentially, what happens looks like this: <P>
544
545The <CODE>hashmap.c</CODE> module tries to detect vertical motion
546changes between the real and virtual screens.  This information
547is represented by the oldindex members in the newscr structure.
548These are modified by vertical-motion and clear operations, and both are
549re-initialized after each update. To this change-journalling
550information, the hashmap code adds deductions made using a modified Heckel
551algorithm on hash values generated from the line contents. <P>
552
553The <CODE>hardscroll.c</CODE> module computes an optimum set of scroll,
554insertion, and deletion operations to make the indices match.  It calls
555<CODE>_nc_mvcur_scrolln()</CODE> in <CODE>lib_mvcur.c</CODE> to do those motions. <P>
556
557Then <CODE>lib_doupdate.c</CODE> goes to work.  Its job is to do line-by-line
558transformations of <CODE>curscr</CODE> lines to <CODE>newscr</CODE> lines.  Its main
559tool is the routine <CODE>mvcur()</CODE> in <CODE>lib_mvcur.c</CODE>.  This routine
560does cursor-movement optimization, attempting to get from given screen
561location A to given location B in the fewest output characters posible. <P>
562
563If you want to work on screen optimizations, you should use the fact
564that (in the trace-enabled version of the library) enabling the
565<CODE>TRACE_TIMES</CODE> trace level causes a report to be emitted after
566each screen update giving the elapsed time and a count of characters
567emitted during the update.  You can use this to tell when an update
568optimization improves efficiency. <P>
569
570In the trace-enabled version of the library, it is also possible to disable
571and re-enable various optimizations at runtime by tweaking the variable
572<CODE>_nc_optimize_enable</CODE>.  See the file <CODE>include/curses.h.in</CODE>
573for mask values, near the end.
574
575<H1><A NAME="fmnote">The Forms and Menu Libraries</A></H1>
576
577The forms and menu libraries should work reliably in any environment you
578can port ncurses to. The only portability issue anywhere in them is what
579flavor of regular expressions the built-in form field type TYPE_REGEXP
580will recognize. <P>
581
582The configuration code prefers the POSIX regex facility, modeled on
583System V's, but will settle for BSD regexps if the former isn't available. <P>
584
585Historical note: the panels code was written primarily to assist in
586porting u386mon 2.0 (comp.sources.misc v14i001-4) to systems lacking
587panels support; u386mon 2.10 and beyond use it.  This version has been
588slightly cleaned up for <CODE>ncurses</CODE>.
589
590<H1><A NAME="tic">A Tour of the Terminfo Compiler</A></H1>
591
592The <STRONG>ncurses</STRONG> implementation of <STRONG>tic</STRONG> is rather complex
593internally; it has to do a trying combination of missions. This starts
594with the fact that, in addition to its normal duty of compiling
595terminfo sources into loadable terminfo binaries, it has to be able to
596handle termcap syntax and compile that too into terminfo entries. <P>
597
598The implementation therefore starts with a table-driven, dual-mode
599lexical analyzer (in <CODE>comp_scan.c</CODE>).  The lexer chooses its
600mode (termcap or terminfo) based on the first `,' or `:' it finds in
601each entry.  The lexer does all the work of recognizing capability
602names and values; the grammar above it is trivial, just "parse entries
603till you run out of file".
604
605<H2><A NAME="nonuse">Translation of Non-<STRONG>use</STRONG> Capabilities</A></H2>
606
607Translation of most things besides <STRONG>use</STRONG> capabilities is pretty
608straightforward.  The lexical analyzer's tokenizer hands each capability
609name to a hash function, which drives a table lookup.  The table entry
610yields an index which is used to look up the token type in another table,
611and controls interpretation of the value. <P>
612
613One possibly interesting aspect of the implementation is the way the
614compiler tables are initialized.  All the tables are generated by various
615awk/sed/sh scripts from a master table <CODE>include/Caps</CODE>; these
616scripts actually write C initializers which are linked to the compiler.
617Furthermore, the hash table is generated in the same way, so it doesn't
618have to be generated at compiler startup time (another benefit of this
619organization is that the hash table can be in shareable text space). <P>
620
621Thus, adding a new capability is usually pretty trivial, just a matter
622of adding one line to the <CODE>include/Caps</CODE> file.  We'll have more
623to say about this in the section on <A HREF="#translation">Source-Form
624Translation</A>.
625
626<H2><A NAME="uses">Use Capability Resolution</A></H2>
627
628The background problem that makes <STRONG>tic</STRONG> tricky isn't the capability
629translation itself, it's the resolution of <STRONG>use</STRONG> capabilities.  Older
630versions would not handle forward <STRONG>use</STRONG> references for this reason
631(that is, a using terminal always had to follow its use target in the
632source file).  By doing this, they got away with a simple implementation
633tactic; compile everything as it blows by, then resolve uses from compiled
634entries. <P>
635
636This won't do for <STRONG>ncurses</STRONG>.  The problem is that that the whole
637compilation process has to be embeddable in the <STRONG>ncurses</STRONG> library
638so that it can be called by the startup code to translate termcap
639entries on the fly.  The embedded version can't go promiscuously writing
640everything it translates out to disk -- for one thing, it will typically
641be running with non-root permissions. <P>
642
643So our <STRONG>tic</STRONG> is designed to parse an entire terminfo file into a
644doubly-linked circular list of entry structures in-core, and then do
645<STRONG>use</STRONG> resolution in-memory before writing everything out.  This
646design has other advantages: it makes forward and back use-references
647equally easy (so we get the latter for free), and it makes checking for
648name collisions before they're written out easy to do. <P>
649
650And this is exactly how the embedded version works.  But the stand-alone
651user-accessible version of <STRONG>tic</STRONG> partly reverts to the historical
652strategy; it writes to disk (not keeping in core) any entry with no
653<STRONG>use</STRONG> references. <P>
654
655This is strictly a core-economy kluge, implemented because the
656terminfo master file is large enough that some core-poor systems swap
657like crazy when you compile it all in memory...there have been reports of
658this process taking <STRONG>three hours</STRONG>, rather than the twenty seconds
659or less typical on the author's development box. <P>
660
661So.  The executable <STRONG>tic</STRONG> passes the entry-parser a hook that
662<EM>immediately</EM> writes out the referenced entry if it has no use
663capabilities.  The compiler main loop refrains from adding the entry
664to the in-core list when this hook fires.  If some other entry later
665needs to reference an entry that got written immediately, that's OK;
666the resolution code will fetch it off disk when it can't find it in
667core. <P>
668
669Name collisions will still be detected, just not as cleanly.  The
670<CODE>write_entry()</CODE> code complains before overwriting an entry that
671postdates the time of <STRONG>tic</STRONG>'s first call to
672<CODE>write_entry()</CODE>, Thus it will complain about overwriting
673entries newly made during the <STRONG>tic</STRONG> run, but not about
674overwriting ones that predate it.
675
676<H2><A NAME="translation">Source-Form Translation</A></H2>
677
678Another use of <STRONG>tic</STRONG> is to do source translation between various termcap
679and terminfo formats.  There are more variants out there than you might
680think; the ones we know about are described in the <STRONG>captoinfo(1)</STRONG>
681manual page. <P>
682
683The translation output code (<CODE>dump_entry()</CODE> in
684<CODE>ncurses/dump_entry.c</CODE>) is shared with the <STRONG>infocmp(1)</STRONG>
685utility.  It takes the same internal representation used to generate
686the binary form and dumps it to standard output in a specified
687format. <P>
688
689The <CODE>include/Caps</CODE> file has a header comment describing ways you
690can specify source translations for nonstandard capabilities just by
691altering the master table.  It's possible to set up capability aliasing
692or tell the compiler to plain ignore a given capability without writing
693any C code at all. <P>
694
695For circumstances where you need to do algorithmic translation, there
696are functions in <CODE>parse_entry.c</CODE> called after the parse of each
697entry that are specifically intended to encapsulate such
698translations.  This, for example, is where the AIX <STRONG>box1</STRONG> capability
699get translated to an <STRONG>acsc</STRONG> string.
700
701<H1><A NAME="utils">Other Utilities</A></H1>
702
703The <STRONG>infocmp</STRONG> utility is just a wrapper around the same
704entry-dumping code used by <STRONG>tic</STRONG> for source translation.  Perhaps
705the one interesting aspect of the code is the use of a predicate
706function passed in to <CODE>dump_entry()</CODE> to control which
707capabilities are dumped.  This is necessary in order to handle both
708the ordinary De-compilation case and entry difference reporting. <P>
709
710The <STRONG>tput</STRONG> and <STRONG>clear</STRONG> utilities just do an entry load
711followed by a <CODE>tputs()</CODE> of a selected capability.
712
713<H1><A NAME="style">Style Tips for Developers</A></H1>
714
715See the TO-DO file in the top-level directory of the source distribution
716for additions that would be particularly useful. <P>
717
718The prefix <CODE>_nc_</CODE> should be used on library public functions that are
719not part of the curses API in order to prevent pollution of the
720application namespace.
721
722If you have to add to or modify the function prototypes in curses.h.in,
723read ncurses/MKlib_gen.sh first so you can avoid breaking XSI conformance.
724
725Please join the ncurses mailing list.  See the INSTALL file in the
726top level of the distribution for details on the list. <P>
727
728Look for the string <CODE>FIXME</CODE> in source files to tag minor bugs
729and potential problems that could use fixing. <P>
730
731Don't try to auto-detect OS features in the main body of the C code.
732That's the job of the configuration system. <P>
733
734To hold down complexity, do make your code data-driven.  Especially,
735if you can drive logic from a table filtered out of
736<CODE>include/Caps</CODE>, do it.  If you find you need to augment the
737data in that file in order to generate the proper table, that's still
738preferable to ad-hoc code -- that's why the fifth field (flags) is
739there. <P>
740
741Have fun!
742
743<H1><A NAME="port">Porting Hints</A></H1>
744
745The following notes are intended to be a first step towards DOS and Macintosh
746ports of the ncurses libraries. <P>
747
748The following library modules are `pure curses'; they operate only on
749the curses internal structures, do all output through other curses
750calls (not including <CODE>tputs()</CODE> and <CODE>putp()</CODE>) and do not
751call any other UNIX routines such as signal(2) or the stdio library.
752Thus, they should not need to be modified for single-terminal
753ports.
754
755<blockquote>
756<code>
757lib_addch.c
758lib_addstr.c
759lib_bkgd.c
760lib_box.c
761lib_clear.c
762lib_clrbot.c
763lib_clreol.c
764lib_delch.c
765lib_delwin.c
766lib_erase.c
767lib_inchstr.c
768lib_insch.c
769lib_insdel.c
770lib_insstr.c
771lib_keyname.c
772lib_move.c
773lib_mvwin.c
774lib_newwin.c
775lib_overlay.c
776lib_pad.c
777lib_printw.c
778lib_refresh.c
779lib_scanw.c
780lib_scroll.c
781lib_scrreg.c
782lib_set_term.c
783lib_touch.c
784lib_tparm.c
785lib_tputs.c
786lib_unctrl.c
787lib_window.c
788panel.c
789</code>
790</blockquote>
791<P>
792
793This module is pure curses, but calls outstr():
794
795<blockquote>
796<code>
797lib_getstr.c
798</code>
799</blockquote>
800<P>
801
802These modules are pure curses, except that they use <CODE>tputs()</CODE>
803and <CODE>putp()</CODE>:
804
805<blockquote>
806<code>
807lib_beep.c
808lib_color.c
809lib_endwin.c
810lib_options.c
811lib_slk.c
812lib_vidattr.c
813</code>
814</blockquote>
815<P>
816
817This modules assist in POSIX emulation on non-POSIX systems:
818<DL>
819<DT> sigaction.c
820<DD> signal calls
821</DL>
822
823The following source files will not be needed for a
824single-terminal-type port.
825
826<blockquote>
827<code>
828alloc_entry.c
829captoinfo.c
830clear.c
831comp_captab.c
832comp_error.c
833comp_hash.c
834comp_main.c
835comp_parse.c
836comp_scan.c
837dump_entry.c
838infocmp.c
839parse_entry.c
840read_entry.c
841tput.c
842write_entry.c
843</code>
844</blockquote>
845<P>
846
847The following modules will use open()/read()/write()/close()/lseek() on files,
848but no other OS calls.
849
850<DL>
851<DT>lib_screen.c
852<DD>used to read/write screen dumps
853<DT>lib_trace.c
854<DD>used to write trace data to the logfile
855</DL>
856
857Modules that would have to be modified for a port start here: <P>
858
859The following modules are `pure curses' but contain assumptions inappropriate
860for a memory-mapped port.
861
862<dl>
863<dt>lib_longname.c<dd>assumes there may be multiple terminals
864<dt>lib_acs.c<dd>assumes acs_map as a double indirection
865<dt>lib_mvcur.c<dd>assumes cursor moves have variable cost
866<dt>lib_termcap.c<dd>assumes there may be multiple terminals
867<dt>lib_ti.c<dd>assumes there may be multiple terminals
868</dl>
869
870The following modules use UNIX-specific calls:
871
872<dl>
873<dt>lib_doupdate.c<dd>input checking
874<dt>lib_getch.c<dd>read()
875<dt>lib_initscr.c<dd>getenv()
876<dt>lib_newterm.c
877<dt>lib_baudrate.c
878<dt>lib_kernel.c<dd>various tty-manipulation and system calls
879<dt>lib_raw.c<dd>various tty-manipulation calls
880<dt>lib_setup.c<dd>various tty-manipulation calls
881<dt>lib_restart.c<dd>various tty-manipulation calls
882<dt>lib_tstp.c<dd>signal-manipulation calls
883<dt>lib_twait.c<dd>gettimeofday(), select().
884</dl>
885
886<HR>
887<ADDRESS>Eric S. Raymond &lt;esr@snark.thyrsus.com&gt;</ADDRESS>
888(Note: This is <EM>not</EM> the <A HREF="#bugtrack">bug address</A>!)
889</BODY>
890</HTML>
891