README.programmer

Parsing:

When reading a file, Mg keeps a count of CRLFs vs. LFs. Biggest one wins
and becomes the buffer mode.

Ditto when reading a file a count is kept of valid UTF-8 characters vs.
other 8-bit characters. Biggest one wins and becomes the mode.

When the counts are equal, a default is used.


Rationale for what charsets are included:

- The Windows OEM codepages from
  http://msdn.microsoft.com/en-us/goglobal/bb964655

- The single-byte Windows codepages from
  http://msdn.microsoft.com/en-us/goglobal/bb964654

- The ISO-8859-* series.

- KOI8-R and KOI8-U.

- UTF-8 - it is the future.

- Arabic and Hebrew are excluded because there is no right-to-left
  display.

There are only ASCII-compatible 8-bit charsets and UTF-8. Other
multi-byte charsets are more tricky to implement and it hasn't been
done.


Suspend and subshell:

There is code for suspending Mg with ^Z where that is supported, but
no code for starting a subshell.


Terminal resizing:

Mg catches SIGWINCH and acts appropriately.


Mods from original Mg:

The ANSI cursor-key sequences are always defined to next-line etc.

Mg always adds a newline to the end of a file if there isn't one. Mg
doesn't even pretend to be a binary editor. For one thing, it doesn't
preserve newlines.


Code guidelines:

Mg3a maintains dot position and buffer size (in bytes). The essential
thing is to not set "curwp->w_dotp" or "curwp->w_doto" directly. Use
adjustpos(), which updates the byte position. Unless you hack
lowlevel, you shouldn't have to update "curbp->b_size".