When adding a new concrete syntax to a grammar, the currently open concrete
syntax will be copied. If the abstract syntax is currently open, the new
concrete syntax will start out empty.
By setting the environment variable GF_RESTRICTED before starting GF, the shell
will be run in restricted mode. This will prevent the GF shell from starting
arbitrary system commands (most uses of System.Cmd.system are blocked) and
writing arbitrary files (most commands that use writeFile et al are blocked).
Restricted mode is intended minimize the potential security risks involved
in allowing public access to the GF shell over the internet. It should be used
in conjuction with system level protection mechanisms (e.g. file permissions)
to make sure that a publicly acessible GF shell does not give access to parts
of the system that should not be publicly accessible.
I changed the version number to 3.2.1-darcs.
I think the version number of the current version from darcs should always
include the "-darcs" suffix (or "-current" perhaps?). The suffix should be
removed only momentarily when making an official release.
This means that the relative grammar directory location in the default
configuration will result in AJAX requests rather than the less efficient
JSONP requests.
Now you can specify more than one mode when building and installing the RGL.
For example to build all modes, run
runhaskell Setup.hs build alltenses present minimal
The default is to build alltenses and present, which means that a single
command,
cabal install
is enough to obtain a normal installation of GF. Without this change, additional
build and install commands would be required before you can compile example
grammars, like Foods and Phrasebook.
The build-binary-dist.sh script has been simplified accordingly.
+ Rename some txt2tags file from .txt to .t2t and remove abandoned .txt files.
+ Add program update_html that finds all .t2t documents and updates the
corresponding .html file. It can be invoked with 'make html'.
+ Add style to some .html documents
If you want the fastcgi server, use cabal install -ffastcgi.
This is to avoid problems when you only want to compile pgf-http and the
fastcgi library fails to install, which it usually does, since it depends on
a C library which is not installed automatically with cabal.
The version of cabal distributed with Ubuntu 10.10 is too old to handle
executables that refer to the library in the same package, so to not make it
too complicated for Ubuntu users to compile GF from source, pgf-http is no
longer part of the main gf.cabal. :-(
The grammars will then by default be fetched from the same server as
minibar.html. This means that everything will be served locally when accessing
minibar through a locally running pgf-http.
Along with the changes to Setup.hs it should now be enough to do
cabal install
pgf-http
open http://localhost:41296/
to run the PGF service locally and play with some example grammars in minibar.
Instead of just giving an explicit list of languages to compile, you can
now also add and remove languages from the default list by using langs=+...
or langs=-...
Examples:
cabal build langs=Eng,Swe,Ger -- as before, compile the given 3 langs
cabal build langs=-Fin -- use default languages, but omit Fin
cabal build langs=+Urd -- use default languages + Urd
("cabal build ..." is the same as "runhaskell Setup.hs build ...")
* The gf command line options -preproc=mkPresent and -preproc=mkMinimal now
refer to internal preprocessors equivalent to lib/src/mkPresent
and lib/src/mkMinimal.
* The temporary file _gf_preproc.tmp is not created when running an
internal preprocessor, unless there is an error, since errors messages
refer to locations in the preprocessed file. (Possibly allowing the rgl
build to be parallelized.)
* After running an external preprocessor, the temporary file is deleted,
unless there was an error.
* (Bug fix) Before, when running more than one preprocessor, the same file name
would be used for both input and output, e.g.,
mkPresent _gf_preproc.tmp > _gf_preproc.tmp
which would result in an empty file being processed. Now, the input and
output files will always be different.
The dependency on the fastcgi package made pgf-server difficult to compile, so
it is now split into
- pgf-fgci (main module in pgf-fcgi.hs), which is built only if fastcgi is
already installed or if you turn on the fastcgi flag (e.g. by doing
'cabal install -f fastcgi').
- pgf-http (main module in pgf-http.hs) which is always built (and hopefully
has no problematic dependencies.)
The modules FastCGIUtils and PGFService no longer depend on fastcgi.
+ PGFService.hs: add Nothing arguments in calls to PGF.parse_ and
PGF.getPArseOutput
+ gf-server.cabal: mark content-server as not buildable, since required source
files are missing.
* Keyboard input and completion should now work much more smoothly:
+ When you press space, the current word will be completed (if incomplete)
and a new magnet will be created. If there is more than one possible
completion, no magnet is created, but the common prefix of the possible
completions is added to the text box.
+ Instead of asking the server for possible completions every time a new
letter is added to the curent word, minibar only ask for completions for
whole words and then filters the list locally when more letters are entered,
speeding things up when server responses are slow.
* Code restructuring:
+ The PGF server API has been moved to its own file: pgf_online.js. This
allows it to be reused in other applicaitons without importing the entire
minibar. It also allows minibar to be used with different server interfaces.
+ The minibar code has been rewritten to avoid storing state information
in the document tree and accessing it by referring to named document
elements. The code now also avoids using string literals contaning
the names of top-level functions to specify event handlers for buttons
and menus. (The code is no longer introspective, so alpha conversion will
not change its meaning.)
+ ajax_http_get(url,callback) now checks the HTTP response status and shows an
error message if the status is not OK (is not 200)
+ New function node(tag,attrs,children) to create document nodes with arbitraty
attributes and children. The functions node() and text() should now be enough
to construct arbitrary document trees in an easy, functional way, and many of
the more specialized functions could be removed.
+ When calling the function button(label,action,key), the action can be a
function instead of a string, so you can use local/anonymous functions.
+ New function debug(str) adds text to the element with id=debug, if it
exists.
Added a meta tag to specify that the file is in UTF-8. This makes the browser
(at least Firefox) assume that the JavaScript files are also in UTF-8, it seems.
Moved the call to start_minibar() from the <body onload="..."> to the end of
the file. The onload action is not executed until all images in the document
has been loaded, while javascipt at the end of the file is executed as soon
as the HTML file has been loaded. (Calling start_minibar() from the head of
the document would be too soon, since it refers to elements in the body,
which hasn't been loaded yet.)
When pfg-server servers a static file in HTTP mode, it is output as is, so
specifying charset="iso-8859-1" could be wrong.
Note: the a charset can be specified with a meta tag in HTML files,
e.g. for files in UTF-8:
<meta http-equiv="Content-type" content="text/html;charset=UTF-8">
hallgren