Write acceptable characters that can be used in --langdef,--fielddef-, --kinddef-, --extradef-, etc.

[masatake]

Though what kind of character we can use as kind letter is written in FORMAT file is not written in FORMAT file , vim accepts only "isalpha" char as kind letter. So ctags should not allow users to define a kind using non-isalpha char. Currently isalnum is used to verify the char used in the letter.

[yamato@master]~/var/ctags-github% ./ctags --kinddef-C='3,three,something three things'
./ctags --kinddef-C='3,three,something three things'
ctags: No files specified. Try "ctags --help".
[yamato@master]~/var/ctags-github% ./ctags --kinddef-C='%,three,something three things'
./ctags --kinddef-C='%,three,something three things'
ctags: the kind letter give in "--kinddef-C" option is not an alphabet or a number

[masatake]

I have to do the same check in --langdef,--fielddef-, --kinddef-, --extradef-, and --regex-=/pat/name/KIND-SPEC/. The man page must be updated, too.
readtags doesn't need such limitation. As far as it can search a tag from given tags file, it should handle any tags file.

[masatake]

• Language names:

  • v2 format spec: nothing specified explicitly. The name is used as a value for "language:" field.
    So the rule for "field value" is applied.
  • u-ctags code for optlib: no check is conducted. However, (Reject more non alnum chars in langdef #2327) u-ctags prints field in escaped format. So even \t is included in the name, the output can be consistent with v3 format.
  • u-ctags man pages: nothing is written.
  • ":-={" these characters should not be used. They break command-line parsing. We have to accept[#+] because they are already used in c++ and c#. Even if tags format allows any characters, it will be better to set limitations in --langdef option handler. How about following categories?
  • [a-z0-9A-Z+#] acceptable
  • [^a-z0-9A-Z+#:-={] warning
  • [:={] error
  • "all" is not allowed. "all" is reserved word. e.g. '--kinds-all=*' enables all kinds of all languages. (Limit to use "all" as the name for a language #2307) optlib2c is not worked yet. (optlib2c: reject "all" as a name of language #2311)

• Kind names:
  Fixing in progress. See [INCOMPATIBLE][SELF INCOMPATIBLE] main,Tmain: revise acceptable characters for kind letters and kind names #1737. Veryfing with Tmain is a must.
  [INCOMPATIBLE][SELF INCOMPATIBLE] main,Tmain: revise acceptable characters for kind letters and kind names #1737 is merged. The man page is readly!

• Field names:

  • v2format spec: alphabetical characters.
  • u-ctags code for optlib: verifying all characters in a field name are alphabetical characters in processLangDefineField.
  • u-ctags code for built-in C crafted parser: no verification. "make tmain" should verify it.
    defineField asserts "islapha". assert macro is enough for detecting a built-in parser introduces wrong field name because ctags on travis-ci is built with --enable-debugging option. The option enables the assert macros.
  • u-ctags man pages: --_fielddef is hidden so no documentation is in the man page.
  • optlib2c rejects other than ^[a-zA-Z]+$.

• Extra names:

  • v2format spec: nothing to do with spec because the name of an extra appears as a value for "extras:" field.
  • u-ctags code for optlib: verifying all characters in a field name are alphabetical or number characters in processLangDefineExtra.
  • u-ctags code for built-in C crafted parser: "make tmain" should verify it.
    defineXtag asserts "isalnum". [a-zA-Z][a-zA-Z0-9]* may be better.(no rational reason).
  • u-ctags man pages: --_extradef is hidden so no documentation is in the man page.
  • optlib2c rejects other than ^[a-zA-Z0-9]+$.

• Role names:

  • v2format spec: nothing to do with spec because the name of an extra appears as a value for "role:" field.
  • u-ctags code for optlib: verifying all characters in a field name are [a-zA-Z0-9] characters in processLangDefineRole. There is no rational reason for this limitation.
  • u-ctags code for built-in C crafted parser: initRoleObject asserts "isalnum".
  • u-ctags man pages: --roledef is hidden so no documentation is in the man page.
  • optlib2c rejects other than ^[a-zA-Z0-9]+$.

[masatake]

About --langdef, everything is done.
The next one is --kinddef. It seems that optlib2c needs some minor work.

[masatake]

--kinddef is mostly done in #2335.
--fielddef is the next one.