| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| |
|
|
|
|
| |
If a name is imported from two places, one hidden and one not, choose
the unhidden one to link to. Also, when there's only a hidden module
to link to, don't try linking to it.
|
| |
|
|
| |
Fix duplicate instance bug
|
| |
|
|
|
|
|
|
|
|
| |
mkExportItems fix & simplification: we should be looking at the actual
exported names (calculated earlier) to figure out which subordinates
of a declaration are exported.
This means that if you export a record, and name its fields separately
in the export list, the fields will still be visible in the
documentation for the constructor.
|
| |
|
|
|
|
|
|
| |
getReExports was bogus: we should really look in the import_env to
find the documentation for an entity which we are re-exporting without
documentation.
Suggested by: Ross Paterson (patch modified by me).
|
| |
|
|
| |
2002 -> 2003
|
| |
|
|
| |
copyright update
|
| |
|
|
| |
getReExports: one error case that isn't
|
| |
|
|
|
| |
Fix for getReExports: take into account names which are not visible
because they are re-exported from a different package.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Re-exporting names from a different package is problematic, because we
don't have access to the full documentation for the entity. Currently
Haddock just ignores entities with no documentation, but this results
in bogus-looking empty documentation for many of the modules in the
haskell98 package. So:
- the documentation will now just list the name, as a link
pointing to the location of the actual documentation.
- now we don't attempt to link to these re-exported entities if
they are referred to by the current module.
Additionally:
- If there is no documentation in the current module, include
just the Synopsis section (rather than just the documentation
section, as it was before). This just looks nicer and was on
the TODO list.
|
| |
|
|
|
|
|
|
|
| |
- Include the OptHide setting in the interface, so we don't include
hidden modules in the combined index/contents.
- Add a -k/--package flag to set the package name for the current set
of modules. The package name for each module is now shown in the
right-hand column of the contents, in a combined contents page.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Add definition lists, marked up like this:
-- | This is a definition list:
--
-- [@foo@] The description of @foo@.
--
-- [@bar@] The description of @bar@.
Cunningly, the [] characters are not treated specially unless a [ is
found at the beginning of a paragraph, in which case the ] becomes
special in the following text.
- Add --use-contents and --gen-contents, along the lines of
--use-index and --gen-index added yesterday. Now we can generate a
combined index and contents for the whole of the hierarchical
libraries, and in theory the index/contents on the system could
be updated as new packages are added.
|
| |
|
|
|
|
|
| |
Include iface_reexported in the .haddock file. This unfortunately
bloats the file (40% for base). If this gets to be a problem we can
always apply the dictionary trick that GHC uses for squashing .hi
files.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Support for generating a single unified index for several packages.
--use-index=URL turns off normal index generation, causes Index
links to point to URL.
--gen-index generates an combined index from the specified
interfaces.
Currently doesn't work exactly right, because the interfaces don't
contain the iface_reexported info. I'll need to fix that up.
|
| |
|
|
|
|
|
|
|
| |
- Suppress warnings about unknown imported modules by default.
- Add a -v/--verbose flag to re-enable these warnings.
The general idea is to suppress the "Warning: unknown module: Prelude"
warnings which most Haddock users will see every time, and which
aren't terribly useful.
|
| |
|
|
|
| |
Remove the last of the uses of 'trace' to emit warnings, and tidy up a
couple of places where duplicate warnings were being emitted.
|
| |
|
|
| |
support for i-parameters + zip comprehensions
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Made -D a short option for --dump-interface.
* Made -m a short option for --ms-help.
* Made -n a short option for --no-implicit-prelude.
* Made -c a short option for --css.
* Removed DocBook options from executable (they didn't do anything),
but mark them as reserved in the docs. Note that the short option
for DocBook output is now -S (from SGML) instead of -d. The latter
is now a short option for --debug.
* The order of the Options in the documentation now matches the order
printed by Haddock itself.
Note: Although changing the names of options is often a bad idea, I'd
really like to make the options for the programs in fptools more
consistent and compatible to the ones used in common GNU programs.
|
| |
|
|
| |
Make it *very* clear that we terminate when given a -V/--version flag
|
| |
|
|
|
|
|
|
|
| |
Made option handling a bit more consistent with other tools, in
particular: Every program in fptools should output
* version info on stdout and terminate successfully when -V or --version
* usage info on stdout and terminate successfully when -? or --help
* usage info on stderr and terminate unsuccessfully when an unknown option
is given.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Rename instances based on the import_env for the module in which they
are to be displayed. This should give, in many cases, better links
for the types and classes mentioned in the instance head.
This involves keeping around the import_env in the iface until the
end, because instances are not collected up until all the modules have
been processed. Fortunately it doesn't seem to affect performance
much.
Instance heads are now attached to ExportDecls, rather than the HTML
backend passing around a separate mapping for instances. This is a
cleanup.
|
| |
|
|
|
|
| |
Pay attention to import specs when building the the import env, as
well as the orig env. This may fix some wrong links in documentation
when import specs are being used.
|
| |
|
|
| |
Update to avoid using hslibs with GHC >= 5.04
|
| |
|
|
|
|
| |
When a module A exports another module's contents via 'module B', then
modules which import entities from B re-exported by A should link to
B.foo rather than A.foo. See examples/Bug2.hs.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
An 80% solution to generating derived instances. A complete solution
would duplicate the instance inference logic, but if a type variable
occurs as a constructor argument, then we can just propagate the derived
class to the variable. But we know nothing of the constraints on any
type variables that occur elsewhere. For example, the declarations
data Either a b = Left a | Right b deriving (Eq, Ord)
data Ptr a = Ptr Addr# deriving (Eq, Ord)
newtype IORef a = IORef (STRef RealWorld a) deriving Eq
yield the instances
(Eq a, Eq b) => Eq (Either a b)
(Ord a, Ord b) => Ord (Either a b)
Eq (Ptr a)
Ord (Ptr a)
(??? a) => Eq (IORef a)
The last example shows the limits of this local analysis.
Note that a type variable may be in both categories: then we know a
constraint, but there may be more, or a stronger constraint, e.g.
data Tree a = Node a [Tree a] deriving Eq
yields
(Eq a, ??? a) => Eq (Tree a)
|
| |
|
|
| |
Small bugfix in the --read-interface option parsing from Brett Letner.
|
| |
|
|
| |
Patches to quieten ghc -Wall, from those nice folks at Galois.
|
| |
|
|
| |
Add a version banner when invoked with -v
|
| |
|
|
| |
merge rev. 1.35
|
| |
|
|
|
|
|
| |
Fix a bug in mkExportItems when processing a module without an
explicit export list. We were placing one copy of a declaration for
each binder in the declaration, which for a data type would mean one
copy of the whole declaration per constructor or record selector.
|
| |
|
|
| |
Sort the options a bit
|
| |
|
|
|
| |
Tweaks to the MS Help support: the extra files are now only generated
if you ask for them (--ms-help).
|
| |
|
|
| |
commented-out debugging code
|
| |
|
|
| |
More bugfixes to the export handling
|
| |
|
|
|
|
|
| |
Clean up the code that constructs the exported declarations, and fix a
couple of bugs along the way. Now if you import a class hiding one of
the methods, then re-export the class, the version in the
documentation will correctly have the appropriate method removed.
|
| |
|
|
|
|
| |
Handle import specs properly, include 'hiding'. Haddock now has a
complete implementation of the Haskell module system (more or less; I
won't claim it's 100% correct).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
When reading an interface, allow a file path offset to be specified
which represents the path to the HTML files for the modules specified
by that interface. The path may be either relative (to the location
of the HTML for this package), or absolute.
The syntax is
--read-interface=PATH,FILE
where PATH is the path to the HTML, and FILE is the filename
containing the interface.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add support for reading and writing interface files(!)
This turned out to be quite easy, and necessary to get decent
hyperlinks between the documentation for separate packages in the
libraries.
The functionality isn't quite complete yet: for a given package of
modules, you'd like to say "the HTML for these modules lives in
directory <dir>" (currently they are assumed to be all in the same
place).
Two new flags:
--dump-interface=FILE dump an interface file in FILE
--read-interface=FILE read interface from FILE
an interface file describes *all* the modules being processed. Only
the exported names are kept in the interface: if you re-export a name
from a module in another interface the signature won't be copied.
This is a compromise to keep the size of the interfaces sensible.
Also, I added another useful option:
--no-implicit-prelude
avoids trying to import the Prelude. Previously this was the default,
but now importing the Prelude from elsewhere makes sense if you also
read in an interface containing the Prelude module, so Haddock imports
the Prelude implicitly according to the Haskell spec.
|
| |
|
|
|
| |
Add support for a "prologue" - a description for the whole library,
placed on the contents page before the module list.
|
| |
|
|
|
|
|
| |
Tiny workaround for the fact that Haddock currently ignores
HsImportSpecs: Let the local_orig_env take precedence.
This is no real solution at all, but improves things sometimes,
e.g. in my GLUT documentation. :-)
|
| |
|
|
|
| |
Keep foreign imports when there is no export list (bug reported by
Sven Panne).
|
| |
|
|
| |
Fix for exporting record selectors from a newtype declaration
|
| |
|
|
|
|
|
| |
Allow exporting of individual class methods and record selectors. For
these we have to invent the correct type signature, which we do in the
simplest possible way (i.e. no context reduction nonsense in the class
case).
|
| |
|
|
|
|
|
| |
Only link to names in the current module which are actually listed in
the documentation. A name may be exported but not present in the
documentation if it is exported as part of a 'module M' export
specifier.
|
| |
|
|
| |
Rename the module documentation properly (bug reported by Sven Panne).
|
| |
|
|
| |
warning message tweak
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Lots of changes:
- instances of a class are listed with the class, and
instances involving a datatype are listed with that type.
Derived instances aren't included at the moment: the calculation
to find the instance head for a derived instance is non-trivial.
- some formatting changes; use rows with specified height rather than
cellspacing in some places.
- various fixes (source file links were wrong, amongst others)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Reworking of the internals to support documenting function arguments
(the Most Wanted new feature by the punters).
The old method of keeping parsed documentation in a Name -> Doc
mapping wasn't going to cut it for anntations on type components,
where there's no name to attach the documentation to, so I've moved to
storing all the documentation in the abstract syntax. Previously some
of the documentation was left in the abstract syntax by the parser,
but was later extracted into the mapping.
In order to avoid having to parameterise the abstract syntax over the
type of documentation stored in it, we have to parse the documentation
at the same time as we parse the Haskell source (well, I suppose we
could store 'Either String Doc' in the HsSyn, but that's clunky). One
upshot is that documentation is now parsed eagerly, and documentation
parse errors are fatal (but have better line numbers in the error
message).
The new story simplifies matters for the code that processes the
source modules, because we don't have to maintain the extra Name->Doc
mapping, and it should improve efficiency a little too.
New features:
- Function arguments and return values can now have doc annotations.
- If you refer to a qualified name in a doc string, eg. 'IO.putStr',
then Haddock will emit a hyperlink even if the identifier is not
in scope, so you don't have to make sure everything referred to
from the documentation is imported.
- several bugs & minor infelicities fixed.
|
| |
|
|
| |
Print the module name in a doc-string parse error
|
| |
|
|
|
|
|
|
|
|
| |
The last commit to Main.lhs broke the delicate balance of laziness
which was being used to avoid computing the dependency graph of
modules.
So I finally bit the bullet and did a proper topological sort of the
module graph, which turned out to be easy (stealing the Digraph module
from GHC - this really ought to be in the libraries somewhere).
|
| |
|
|
| |
Add support for existential quantifiers on constructors.
|
