otags(1) generate tags files for emacs and vi/vim from OCaml sources

SYNOPSIS

otags [STATIC-OPTION] ... [FILE] ... [DYNAMIC-OPTION] ... [FILE] ...

DESCRIPTION

otags generates TAGS files for emacs and tags files for vi/vim to give these editors the ability to directly jump to function and value definitions (use M-. in Emacs and ^] in vi[m] with the cursor on the symbol).

otags uses camlp4 for parsing. This is more precise than regular expression based tagging. On the downside, otags can only tag syntactically correct files. Furthermore, otags is OCaml version specific: Sources to be compiled with OCaml version X.Y can, in general, only be tagged with otags version X.Y.

otags can tag all files that camlp4 can parse. Internally otags keeps a current parser list that can be modified with the options -pc (clear list), -pa (add to list), and -pr (reset to default). The default value of the current parser list corresponds to standard OCaml syntax (containing the parsers r and o, see STANDARD CAMLP4 PARSING EXTENSIONS below). Each source file is parsed with a camlp4 parser into which all parsing extensions from the current parser list have been loaded. The camlp4 standard parsers are linked into otags and if the current parser list contains only such parsers then the parsing module is build internally. Otherwise an external camlp4 process is started.

otags tags all toplevel items in structures and signatures, including infix operators and nested modules and signatures. It also tags instance variables and methods in classes and class-types. Filenames FILE on the command line are parsed as interfaces if they have a .mli suffix and as implementations if they have a .ml suffix. (See also the dynamic options -intf and -impl.)

otags can tag all files in a directory tree (if option -r is given). Files that are generated by ocamllex or ocamlyacc or that require preprocessing with the camlp4 macro parser must, however, be tagged in the same directory where these tools have been run. This is because the line directives that ocamllex and ocamlyacc put into generated files are relative to the current working directory of the generating process. For the camlp4 macro parser an INCLUDE directive may be resolved relative to the current working directory of the camlp4 process. Therefore, if you tag such files from a different directory you will get an error message.

When tagging whole directory trees recursively, one can specify exceptions from the current parser list in a parser hints file or directory tree, see option -parser-hints and PARSING HINTS FILES below.

Tag files can contain absolute and/or relative file names. Otags uses the file names from the command line for the tags file. Specifying relative file names on the command line will therefore give you a tag file with relative file names. Obviously, a tag file with relative file names will only work if it is located in the directory where otags was started. Otags supports writing a tags file with relative file names in a parent directory via option -add-path, see EXAMPLES below.

STATIC OPTIONS

Static options take effect on the whole program run, regardless of their position in the command line.
-r
Descend into directories. With this option, if one of the FILE arguments is a directory otags will recursively tag all files in this whole directory tree.
-o file
Write tags to file. The default is TAGS (for emacs) and tags for vi (if the -vi option is given) in the current directory. If file is a dash ( - ) the tags are written to standard output.
-a
Append to an existing TAGS file (does only work for emacs TAGS files).
-vi
Generate tags for vi and change the default output file into tags.
-I dir
Add directory dir to the camlp4 search path for object files for external camlp4 parsing calls.
-add-path path
Add path at the front of every relative file name in the tags file. Useful when you want to put the tags file in a parent directory.
-parser-hints file
Apply the parser hints in the specified file or directory tree. Whenever otags tags a file that appears in a parser hints file, it uses the parsers specified there instead of the current parser list, see PARSING HINTS FILES below. This option can be given multiple times to process several hints files. If file is a directory then all files in that tree are processed as parser hints files.
-version
Print version and exit.
-v
Verbose. Print file names as they are parsed and the command lines for externally started camlp4 parsing processes (if any). If -v is the first command line argument backtraces for exceptions are enabled and printed if an exception escapes and terminates otags (which is considered as a bug about which I would appreciate a bug report).
-q
Be quiet.
-help
Print the option list.
--help
Alias for -help

DYNAMIC OPTIONS

Dynamic options do only affect arguments that follow them on the command line.
-intf FILE
Parse and tag FILE as an interface.
-impl FILE
Parse and tag FILE as implementation.
-pc
Clear the current parser list.
-pa parser
Add parser to the current parser list. If parser is not one of the standard parsers that are distributed with camlp4 then the following file(s) are parsed in separate camlp4 process(es). parser must be accepted by camlp4 -parser, that is, it must be a standard camlp4 parser name or a bytecode file (.cmo) or library (.cma).
-pr
Reset the current parser list to its default value (parse standard OCaml without extensions).
-pp
Print the current parser list.
-extern
Force otags to parse all the following files with an external camlp4 process.
-intern
Switch back to (default) internal parsing.

PARSING HINTS FILES

Parsing hints files are simple text files that specify parsers for certain files. This is useful when tagging recursively with option -r. Here is an example for some files from the OCaml distribution:

oq:
        camlp4/examples/parse_files.ml
o rq g:
        camlp4/examples/macros.ml
        camlp4/examples/arith.ml
rf debug:
        camlp4/Camlp4Bin.ml
        camlp4/mkcamlp4.ml
        camlp4/camlp4prof.ml
        camlp4/camlp4prof.mli

The format of parser hints files is as follows. Empty lines and lines starting with ``#'' are ignored. A parser hints file consists of an arbitrary number of sections. Each section starts with a colon terminated line that specifies the parser to use with a space separated list of parser names (like those accepted by option -pa). The remainder of the section contains file names, one per line. Leading and trailing white space is ignored.

EXAMPLES

Tag some files in the original and some in the revised syntax:

otags original.ml -pc -pa r revised.ml -pr original.ml

The -pc is necessary because, for camlp4, the original OCaml syntax is an extension of the revised syntax. Therefore, the default current parser list contains the parser r already and adding r would not change the current parser list.

Tag files with quotations (original host syntax and revised quotation syntax):

otags -pa rq qotation.ml

Tag files that use the sexplib extension:

otags -I /usr/lib/ocaml/type-conv -I /usr/lib/ocaml/sexplib -pa pa_type_conv.cmo -pa pa_sexp_conv.cmo use_sexp.ml

The -I options specify the directories where pa_type_conv.cmo and pa_sexp_conv.cmo are located on the system.

Tag files in a subdirectory keeping the tags file in parent directory

cd subdir; otags -o ../TAGS -add-path subdir ...

This complicated procedure is only necessary if there are files inside subdir with line directives containing relative file names (for instance, generated by ocamllex or ocamlyacc running inside subdir) and when there are other sudirectories whose material should be added to ../TAGS. If there are no files with relative line directives one can simply use otags -r in the directory containing subdir. If there are no other subdirectories one could put the tags file into subdir without using -o or -add-path.

STANDARD CAMLP4 PARSING EXTENSIONS

There is currently no documentation on the names and aliases that refer to standard camlp4 parsing modules. Also the dependencies among those modules is mostly undocumented. The only source for information is the file camlp4/Camlp4Bin.ml in the OCaml distribution.

The following list describes the identifiers that camlp4 -parser treats specifically. This is therefore the list of standard camlp4 parsers. For these identifiers case is not significant (they are piped through String.lowercase before matching). The dependencies listed here are treated auto-magically by camlp4 and otags.

r, ocamlr, ocamlrevised, pa_r.cmo, camlp4ocamlrevisedparser.cmo
Revised syntax without stream parsers.
rr, reloaded, ocamlreloaded, camlp4ocamlreloadedparser.cmo
Variant of the revised syntax (usage unknown).
o, ocaml, pa_o.cmo, camlp4ocamlparser.cmo
Original syntax without stream parsers. Depends on parser r.
rp, rparser, pa_rp.cmo, camlp4ocamlrevisedparserparser.cmo
Revised syntax with stream parsers. (See bug #5134 for OCaml 3.12.0 and earlier.) Depends on parser r.
op, parser, pa_op.cmo, camlp4ocamlparserparser.cmo
Original syntax with stream parsers. Depends on parser r, o, rp
g, grammar, pa_extend.cmo, pa_extend_m.cmo, camlp4grammarparser.cmo
Grammar extensions.
m, macro, pa_macro.cmo, camlp4macroparser.cmo
Macros and conditionals.
q, camlp4quotationexpander.cmo
Reflective Quotations. This parsing extension takes the current host syntax and adds quotation expanders using that current syntax such that the resulting grammar is reflective: Changing the host syntax afterwards does also affect the quotation syntax. Depends on module Camlp4QuotationCommon (common quotation infrastructure).
rq, q_mlast.cmo, camlp4ocamlrevisedquotationexpander.cmo
Revised Quotations. Adds quotations in the revised syntax. Host and quotation parsing remain independent. Depends on module Camlp4QuotationCommon (common quotation infrastructure).
oq, camlp4ocamloriginalquotationexpander.cmo
Original OCaml with original quotations. Loads original OCaml (without stream parsers) as host syntax and as quotation syntax. Host and quotation syntax are independent (not reflective). Depends on parser ocamlr, ocaml and module Camlp4QuotationCommon
rf
Full revised. Sets up revised syntax with stream parsers, grammar extensions, list comprehensions and macros and adds reflective quotations. Depends on parsers/modules r, rp, Camlp4QuotationCommon, q, g, comp, m.
of
Original full. Sets up original OCaml with stream parsers, grammar extensions, list comprehensions and macros and adds reflective quotations (i.e., quotations are in original syntax). (See bug #5129 for OCaml 3.12.0 and earlier.) Depends on parsers/modules r, o, rp, op, Camlp4QuotationCommon, q, g, comp and m.
comp, camlp4listcomprehension.cmo
List comprehensions.
debug, camlp4debugparser, camlp4debugparser.cmo
Debugging statements debug and camlp4_debug. The expansion of those statements depends on the value of the environment variable STATIC_CAMLP4_DEBUG such that also the produced tag entries might depend on this variable.

DIAGNOSTICS

Parse errors are reported with normal camlp4 error messages. No tags are generated for files that produce errors. Tagging continues with the next file, in this case.

The error "Original source not available" is reported if the file to which a line directive refers cannot be found. This may happen if otags is started in a different directory than the program that generated the faulting file.

Otags exists with "Parser list empty for ..." when attempting to process a file with an empty current parser list. This happens when no -pa option follows -pc.

Otags prints "Toplevel directive found in ... Skip file." for files containing toplevel directives. The problem is that the location information in the abstract syntax tree is incorrect after toplevel directives, see bug #5127.

Escaping exceptions and assertions that terminate otags are considered as bugs. If they occur, please use option -v as first command line switch to obtain an exception backtrace and submit this with all relevant information as bug report.

EXIT STATUS

0
everything went OK
1
a parse error or some other error occurred and otags skipped the affected file
2
a fatal error caused otags to abort
3
some bug caused a crash, from which otags could only marginally recover

KNOWN PROBLEMS

The camlp4 and OCaml parsers differed always in subtle ways. Making camlp4 independent of OCaml aggravated the problem. camlp4 version 4.02.3 only accepts very few attributes. It accepts no block or stand alone attributes (those with [@@ and [@@@) or extension nodes ( [% or [%%). Otags therefore fails on many fails from the OCaml standard library.

MISSING FEATURES

Appending to vi tags files requires to reread the existing tags file, because vi tags files are sorted.

Incompatible camlp4 syntax trees (from external camlp4 parsing processes) are detected with magic numbers. However, an incompatible camlp4 will likely die when it tries to load the otags specific camlp4 printer. Such fatal dynamic loader errors are treated like parsing errors, because camlp4 produces the same exit status in both cases.

CREDITS

Cuihtlauac Alvarado and Jean-Francois Monin were the first to exploit camlp4 for tagging OCaml files. This version is a complete rewrite, based on the new camlp4 from OCaml version 3.10 and onwards.

AUTHOR

Hendrik Tews <otags at askra.de>