3e1db26a5a
Removing elvis, importing nvi, ctags, updating libedit. Change-Id: I881eb04d2dc64cf112facd992de1114e1a59107f |
||
---|---|---|
.. | ||
dump.c | ||
dutch.base | ||
dutch.owner | ||
english.owner | ||
french.base | ||
german.base | ||
german.owner | ||
Makefile | ||
README | ||
ru_RU.KOI8-R.base | ||
ru_RU.KOI8-R.owner | ||
spanish.base | ||
spell.ok | ||
swedish.base | ||
swedish.owner |
# Id: README,v 8.4 1994/11/22 09:52:04 bostic Exp (Berkeley) Date: 1994/11/22 09:52:04 Generally, all non-system error and informational messages in nvi are catalog messages, i.e. they can be tailored to a specific langauge. Command strings, usage strings, system errors and other "known text" are not. It would certainly be possible to internationalize all the text strings in nvi, but it's unclear that it's the right thing to do. First, there's no portable way to do message catalogs. The System V scheme is a reasonable choice, but none of the 4BSD derived systems support it. So, catalogs are completely implemented within nvi, and don't require any library support. Message catalogs in nvi are fairly simple. Every catalog message consists of two parts -- an initial number followed by a pipe (`|') character, followed by the English text for the message. For example: msgq(sp, M_ERR, "001|This is an error message"); would be a typical message. When the msgq() routine is called, if the user has specified a message catalog and the format string (the third argument) has a leading number, then it is converted to a record number, and that record is retrieved from the message catalog and used as a replacement format string. If the record can't be retrieved for any reason, the English text is displayed instead. Each message format string MUST map into the English format string, i.e. it can't display more or different arguments than the English one. For example: msgq(sp, M_ERR, "002|Error: %d %x", arg1, arg2); is a format string that displays two arguments. It is possible, however, to reorder the arguments or to not display all of them. The convention nvi uses is the System V printf(3) convention, i.e. "%[0-9]*$" is the name of a specific, numbered argument. For example: msgq(sp, M_ERR, "002|Error: %2$d %1$x", arg1, arg2); displays the arguments in reverse order. If the system supports this convention in its library printf routines (as specified by the test #define NL_ARGMAX), nvi uses those routines. Otherwise, there is some serious magic going on in common/msg.c to make this all work. Arguments to the msgq function are required to contain ONLY printable characters. No further translation is done by the msgq routine before displaying the message on the screen. For example, in the msgq call: msgq(sp, M_ERR, "003|File: %s", file_name); "file_name" must contain only printable characters. The routine msg_print() returns a printable version of a string in allocated memory. For example: char *p; p = msg_print(sp, file_name); msgq(sp, M_ERR, M("003", "File: %s"), p); FREE_SPACE(sp, p, 0); makes sure that "file_name" is printable before calling the msgq routine. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= The message catalogs themselves are maintained in two files. The first is the "base file" which contains two fields, a record number and the message itself. All base files are named using the convention "vi_<language>.base", e.g. the English one is "vi_english.base". For example: 002 "Unable to create temporary file" 003 "Warning: %s is not a regular file" 004 "%s already locked, session is read-only" 005 "%s: remove" 006 "%s: close" 007 "%s: remove" 008 "%s: remove" 009 "Read-only file, not written; use ! to override" 010 "Read-only file, not written" are the first few lines of the current vi_english.base file. Note that message #1 is missing -- the first message of each catalog is a special one, so that nvi can recognize message catalog files. It's added by the Makefile script that creates the second version of the message catalog. The second file is the file used by nvi to access messages, and is a list of the messages, one per line: VI_MESSAGE_CATALOG Unable to create temporary fileX Warning: %s is not a regular fileX %s already locked, session is read-onlyX %s: removeX %s: closeX %s: removeX %s: removeX Read-only file, not written; use ! to overrideX Read-only file, not writtenX Note that all messages have had a trailing 'X' character appended. This is to provide nvi a place to store a trailing nul for the message so that C library routines that expect one won't be disappointed. These files are named for their language, e.g. "vi_english". The second files are automatically created from the first files. To create a new catalog for nvi: Copy the file vi_english.base to a file that you can modify , e.g. "cp vi_english.base vi_german.base". For each of the messages in the file, replace the message with the string that you want to use. To find out what the arguments to a message are, I'm afraid you'll have to search the source code for the message number. You can find them fairly quickly by doing: cd ..; egrep '123\|' */*.[chys] I'm sorry that there's not an easier way, but I couldn't think of anything that wasn't a lot of work. If, for some reason, you don't have the file vi_english.base, or you have new sources for which you want to create a new base catalog, you can create it by running the command "make english" in the catalog directory. Once you've translated all of the strings, then add your catalog to the "CAT=" line of the Makefile, and run the command "make catalog". This will create the second (and corresponding) file for each file named <language>.base. Don't worry about missing line numbers, i.e. base files that look like: 005 Message number 5. 007 Message number 7. This simply means that a message was deleted during the course of nvi's development. It will be taken care of automatically when you create the second form of the file. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= If you add new messages to the nvi sources, you can check your work by doing "make english; make check". The "make check" target lists unused message numbers, duplicate message numbers, and duplicate messages. Unused message numbers are only useful if you are condensing messages. Duplicate message numbers are a serious problem and have to be fixed. Duplicate messages are only interesting if a message appears often enough that it's worth creating a routine so that the string is only need in a single place. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= To select a catalog when running nvi, set the "msgcat" option. If the value of this option ends with a '/', it is treated as the name of a directory that contains a message catalog "vi_XXXX", where XXXX is the value of the LANG environmental variable, if it's set, or the value of the LC_MESSAGES environmental variable if it's not. If neither of those environmental variables are set, or if the option doesn't end in a '/', the option is treated as the full path name of the message catalog to use. If any messages are missing from the catalog, the backup text (English) is used instead.