- Wish automatic (optional) backup on write.
e.g. attribute backup specifies the extension to be used
for a backup file. If present and non-null, and the file to be written
already exists, and the append attribute is not present,
the existing file is renamed to create a backup file. If the value of
backup ends with a period, it is inserted before
the filename extension, otherwise it is simply appended. If the value
contains the character ``0'' (zero), it is replaced with
the smallest integer for which a corresponding backup file does not
already exist.
(2000-12-05)
-- non-trivial because <output> uses locate,
locateSystemResource, etc. May need "backup" parameter
to locateSystemResource and documentWriter.
- Hole streaming lazy evaluation of entities would be nice.
Currently, expanding an EntityRef involves constructing an intermediate
NodeList even if the EntityRef could provide an Input. This is
because there's no way for an Input to perform processing. It might be
sufficient for a Processor to expand entities on its input by pushing
them onto an input stack.
(2000-10-05)
- Hole maintain tagset dependencies
invalidate timestamps of dependents when reloading a tagset.
(2000-07-20)
- Hole need a different Output for HTML
handling for optional end tags and empty delimiters is OK for XML
(they're ignored or smashed when appropriate), but not for HTML)
(2000-03-07)
-
Release 2.1.1
- Check use of nodeValue.
Should probably be eliminated; used in Util/EntityInput and
namespace/EntityWrap.
(2000-02-25)
- Impr Get attribute-list case sens. from tagset
Need to case-fold attributes correctly for the tagset.
- Impr
syntax="properties"
extension for define: attributes and sub-elements go into
the local (stack frame) namespace.
- Check TopProcessor may no longer need prefix support.
should do it via the root. May still be needed for "process"
(1999-12-14)
- Check verify correct content model for HTML
The stuff in HTML_ts needs to be checked against the official DTD.
(1999-12-03)
- Bug New attributes of <date> undocumented.
(1999-11-17)
- Bug Storing a Namespace in a Namespace needs an entity
But on the other hand, storing a list in a PropertyTable needs an
element. Really needs to be handled in addBinding.
(1999-11-17)
- Bug Entity def's aren't available in a tagset
Therefore set's, etc. won't work there.
(1999-11-11)
- Bug Passive entities don't seem to be recognized.
(1999-11-11)
- Bug Entity expansion shouldn't expand passives
Right now it uses Expand.expandNodes, which doesn't check for
passives. Converting to internal form does
expand passives, of course.
(1999-11-11)
- Impr Delete TreeExternal.close*put.
ConnectHandler should just close the Input or Output. Requires a
check.
(1999-11-04)
- Impr include using document's own tagset
There should be an option to include that allows a document to use
its own tagset.
(1999-09-13)
- Hole Replace File with Resource almost everywhere
One known example is util.Forms where it wants a file's MIME type.
Another more egregious one is <include>!
(1999-09-13)
-- (DONE: include) status ...
- Impr define !DOCTYPE for XML
Should define the !DOCTYPE node in the default XML tagset with an
action that switches to the correct tagset as specified by the
doctype.
(1999-09-13)
- Hole locate*Resource now problematic
Should end up with a Resource rather than a File or URL. A
lot of things would be simpler if we used Resource, etc. for
all access to resources.
(1999-09-07)
- Bug <replace node> buggy
Needs to replace with whole content, not first node in content.
By the way, node isn't documented in basic.ts.
(1999-08-09)
- Impr <bind type="namespace">
Would make it possible to set up sub-namespaces at load time.
(1999-07-15)
- Bug <include src="file:..."> fails for directories
There also ought to be a way to get directory content from
<status> [there is: item="files"].
(1999-07-14)
-- This is because readExternalResource just blindly returns a
FileInputStream instead of going through FileAccess (PIA) or the
equivalent (Filter).
- Bug <connect> gives null pointer exceptions.
(1999-07-14)
- Impr safeDefineHandler, safeSetHandler
Safe versions of <define> and <set> that are only able to
define things in the current document, and with no handlers. There
should also be a <make-safe> tag to use safe versions of all
handlers. Might also want a "sandbox" subclass of ActiveDoc that
effectively makes the AGENT, PIA and tagset namespaces read-only.
(1999-07-14)
- Impr Need action code PROCESS_NODE
This would indicate an active node with start and end actions that can
be invoked without having to retain the node's content. Need
processStart and processEnd ops on Processor; something that define can
use as well.
(1999-07-14)
- Hole ToProcessor untested.
(1999-07-14)
- Bug <text encode> broken on markup
It's removing all markup; for sanity, should convert to external form.
May also not be expanding content.
(1999-07-12) [Bill]
- Impr <include quoted> -> protect
It's inconsistent to use "quoted" here and "protect" elsewhere. Could
even have "protect markup" sucking in the file as a string
or list of lines.
(1999-07-02) [Suggested by Bill]
- Hole Need split and join attributes on text, etc.
Anything that processes lists should have "split" and "join"
attributes, replacing the old "sep". Keep "sep" as synonym for
"join", which it is.
(1999-06-29)
- Bug nodeBuilder doesn't support all node types yet
(1999-06-24)
- Bug Bad end tag comment sometimes eaten.
Actually, if it's not quoted it seems to be replaced by a copy of the
previous token.
(1999-06-16)
-- May have to do with a missing hasChildren test.
- Impr Parser error verbosity level
The verbosity level of parser messages should be a tagset
parameter: for XML they should be warnings or perhaps errors; for HTML
most if not all should be verbose. Similarly treatment of ommitted end
tags and empty tags should be a parameter related to the output's DTD.
(1999-06-16)
- Bug Null pointer error in Utilities.urlDecode
org.risource.util.Utilities.urlDecode(Utilities.java:729) called from
text_decode.action(textHandler.java:476) on request
for http://blackroot.crc.ricoh.com:8888/Demo/demo_text.xh
(1999-06-14 Pfafman)
- Bug <text encode> badly broken
Should be <text encode=entity> not <encode entity>.
Also, it seems to be doing bad things to an entity in its contents, in
some totally incomprehensible way.
(1999-06-08)
- Impr allowing text in attr lists preserves formatting
It would also allow the use of entities that expand to multiple
attributes.
(1999-05-25)
- Bug <bind> doesn't build external entities
For that matter, <bind> has no way to build any value
other than a list of ordinary nodes. Might actually want to generate
<define> for nonstandard bindings. Or <binding>...
(1999-05-25)
- Bug Need way to sync an external entity.
Presumably some version of <connect>, the idea being to write the
current cached contents of the entity node. Wanted for, e.g.,
agent-list.xml. Actually, <output> may be simpler.
(1999-05-24)
- Hole need input redecorator
needed for TopProcessor.subDocument and subDocumentEnd when the input
is not a parser.
(1999-05-19)
- Hole XML Attribute handling
Almost everything currently done in the PIA using the Syntax interface
could be done using inherited attributes. For example, there's an
xml:space attribute specifying whitespace handling.
(1999-05-17)
- Hole writeup of <test> seriously incomplete.
For some reason it had been omitted from the table in
/Doc/tagsets/basic.html!
(1999-05-17)
- Bug <bind> doesn't build embedded <namespace>
Because <bind> is quoted and <namespace> doesn't construct
a namespace at parse time, this simply won't work. Could actually emit
a <namespace> tag in this case.
(1999-05-14)
- Impr Drop support for serialized objects
It turns out that loading XML is faster, at least for stripped tagsets.
Any further improvements to XML processing will only widen the gap.
(1999-05-07)
- Impr XML Representation of Tagset.
Use <ELEMENT>, <ENTITY>, <bind> tags in content;
construct the objects in the parser(?). Save the flags, etc. as
attributes. Dump the *.tso files in favor of
.tsx.
(1999-05-07)
- Impr Namespace attribute in get, set.
This would allow the name to be literally any text. Would require
minor fixes in Index to allow ``:'' in namespace arg, not name.
(1999-05-03)
- Impr xxml additional constructs.
Eventually xxml should include stuff for, e.g., reading
objects and building data structures.
(1999-05-03)
- Hole Need actions for PI's and Decl's
In particular, an action on <!DOCTYPE....> would allow
both runtime selection of tagset, and separate
specification of the output document type.
(1999-05-03)
- Impr Output to Writer (or ToPrint) not String.
Something like writeNode(Writer); writeStart, writeContent, writeEnd
ops for ActiveElement. ActiveNode.toString can just use it. The
start, content, end versions would have parameters (e.g. indent).
Alternative is keeping context in a specialized Output.
(1999-04-29)
- Impr NamespaceWrap may need to use EntityIndirect.
(1999-04-16)
- Bug Output to file in non-existant directory fails
It should create the directory; instead it gets a null pointer
exception in Copy.copyNodes, called from
connectHandler.action.
(1999-04-20)
-- fixed in GenericAgent 4/21. process will still have the problem.
- Bug Entities don't have system and public ident's yet
In particular, TreeExternal needs these.
(1999-04-20)
-- resourceName == systemId (4/28). Still need creation machinery.
- Impr Get rid of util.Create
It's only used in three places. All uses of createActiveNode should be
replaced by type-specific handler classes in Tagset. None of the
nodelist-creation methods are used anymore.
(1999-04-20)
- Bug Tagset should extend DocumentType
It's completely clear that in the DPS world, Tagset has all the
attributes of a DocumentType object.
(1999-04-20)
- Impr fix inheritance problems; simplify.
ActiveEntityRef shouldn't descend from ActiveValue, since it doesn't
have one. Namespace should extend, if not merge with, ActiveNodeMap.
util.Expand should probably go away. ProcessException should become
a DPSException. Case stuff should move out of Namespace.
(1999-04-07)
ActiveEntityRef fixed 4/15. ProcessException turns out to be unused;
deleted 4/16.
- Hole Fully-cursorial processing.
Cursor ought to give complete enough access to the current node that
you never actually need to access the node itself. ActiveNode needs to
be extended with methods that return Input's instead of Node's. Output
needs Document-like methods to output nodes without having to create
them first.
(1999-04-07)
- Hole unimplemented DOM components
CDATASection is unimplemented. DOMImplementation is unimplemented.
Large parts of TreeDocument are still unimplemented; in particular
extracting the DocumentType and document element nodes from the content
and putting them in their proper places.
(1999-04-07)
-- CDATASection done 1999-06-24
- Hole SAX and DOM input and output
FromDOM, ToSAX, ToDOM shims. Note that FromDOM can be used to
redecorate a parse tree with new semantics. (From SAX requires an
event-driven Processor).
(1999-04-07)
- Bug <include>Directory broken
<include src='~/'> should give a directory listing of an agent's
data directory. It raises a null pointer exception instead.
(1999-03-30)
- Impr SGML/XML output flag
There needs to be a flag somewhere that says whether to output as SGML
or XML.
(1999-03-30)
- Impr Need ways of getting at node names, types, keys
Mainly in <extract>, though a general mechanism would be good.
It's possible that the empty <extract> sub-elements could be
generalized so that they operate on their content, if present, instead
of on &list;.
(1999-03-29)
- Impr Really need <action mode=prepend>
The action should be able to edit the attributes of &element;, but
the content would then be streamed through.
(1999-03-11)
- Bug <value> ignored in <define entity external>
The value sub-entity is ignored when using the define entity system
attribute. For example, the value "whatever" never gets written to file.:
<define entity=pgroup system=per-group mode=create><value>whatever</value></define>
&pgroup;
(1999-02-23)
- Bug Test atributes not all given in
basic.ts
Also, the "sorted" attribute appears to be buggy. Perhaps it is doing
a numeric sort instead of an alphabetic one, or not doing a split.
(1999-02-17)
- Bug &agents; is static, not dynamic
...so the header is wrong when agentRemove or agentInstall is used.
This is only a problem in Agency, so we can make AGENT:agents return
the up-to-date list.
(1999-02-02)
- Hole Namespaces for tags
Tags need namespaces. The namespace defined by a tagset might not be
the same as its name; wellknown namespaces will include:
-
dps:
- all tags defined in the
basic, xxml,
and xhtml tagsets
-
pia:
- all tags defined in the
pia-xhtml tagset.
(1999-02-02)
- BugConnect
When doing a "connect put" cannot specify src="/RHServer/persist.out"
from an agent. It throws an exception. src="persist.out" does not
throw an exception, but it does fail to write output to file. The same
code run using process works fine. "connect entity" requires
"/RHServer/persist.out" when running from an agent.
(1999-01-26)
- Hole Data structure elements
<headers>, <url>, etc. need to convert text content to XML
format, and need to output as text.
(1998-12-14)
- Hole list elements as namespaces
Need to support two things: list tags (e.g. <dl>) as namespaces,
and entities recognizing them when their value contains a
namespace or name. Would be easier if there was a simple way for the
parser to construct arbitrary classes for elements.
(1998-12-04)
- Impr Persistent namespaces
There are actually two ways to make a namespace persistent: make it
the value of an external entity, and make its bindings be external
entities. In the latter case, cons up filenames from the entity
names with a prefix.
(1998-11-24)
- Note Tagset for external entity.
Note that an external entity can go up its parent chain to find its
tagset. If not given a context explicitly, it can make its own
TopProcessor when needed. Note that external entities do want a
tagset; it's basically the notation.
(1998-11-19)
-- The problem here is that the entity needs to know what TopProcessor
to use in order to open files; this needs to be passed at run time.
- Note Tagset should be able to create a Processor.
They can already create their parser. There are times when you want a
different processor type, but in most cases the right thing will happen.
(1998-11-19)
- Impr Resource interface.
Everything having to do with external resources (entities, connections,
and status) would be much simpler if we had a Resource interface.
Something like URLConnection only more versatile.
(1998-11-18)
- Bug Assigning attributes to an entity should make namespace.
Assigning an actual AttributeList works, but using <set>
doesn't. Probably best to fix this in setValue.
(1998-11-18)
- Impr Pipelined <extract>
Should be possible to stream through a series of processors. Bind
&item; to current item; evaluate &selection; lazily as needed.
(1998-10-21)
- Impr shorthand for booleans in extract.
equivalent to test but with current item in content.
(1998-10-21)
- Bug Test sorted. Returns true on an unsorted list. For example:
Input: <test sorted>bats ants toads</test>
Result: 1
(1998-10-19)
- Bug Pad only pads on the right (align left). For example:
Input: <pre><text pad align=center width=80>cat</text></pre>
Results: <pre>cat </pre>
(1998-11-11)
- Impr Prevent infinite loops. Here's one way to create one:
<repeat foreach entity=x>a b c d e f g><repeat><for><start>2</start><stop>5</stop>&x;</for></repeat></repeat>
(1998-10-19)
- Bug: subst has trouble with matches that get to the end.
This is a generic problem with the regexp stuff.
(1998-10-9)
- Bug: literal element not properly terminated.
(1998-10-2)
- Bug: addMarkup has to insert real markup now.
...because Text now converts markup characters to entities.
(1998-10-2)
- Impr: Use match to split on regexp.
<text split match=re> should convert content to a single
string, then split on the given regexp. This solves the problem of
how to make sep consistent between split and
join.
(1998-09-15)
- Hole: Some <extract> subs unimplemented.
xptr is the only one left.
(1998-09-08)
-- this can easily be deferred
- Hole: Parser work needed.
Literals, CDATA sections, parent, content models, declaration and PI
parsing, optionally splitting text into tokens, optional case-munging.
(1998-08-13)
-- retainTree implemented 08-27 but untested.
parent stuff (no error checking) 09-11. literals 09-18/10-1.
- Hole: Attribute definitions.
Attribute types, implied/fixed/etc., default values, handlers.
Dispatch table for attrs with handlers. Lazy evaluation(?)
(1998-08-13)
- Hole: Text Substitution
Needs to be done using either <select> or its own <subst>
element with appropriate sub-elements.
(1998-08-12)
-- (1998/08/26) handler written but no sub-elements yet.
- Bug: Tagsets don't hack resource names correctly.
Need to decide between Class and ClassLoader conventions. ClassLoader
requires slashes; Class prepends module name. Need the prepend, then
change dots to slashes (probably in
org.risource.util.NameUtils).
(1998-08-12)
- Bug: Can't use a tagset.
The <tagset> tag only really works in a tagset definition file.
(1998-08-11)
-- Possibly best done using a separate useTagset handler.
- Bug: Handler class hierarchy problems.
There are two many places where we assume that all real handlers
inherit from BasicHandler. Fix one way or the other.
(1998-08-13)
- Impr: Functional programming tools.
See Python
(1998-08-05)
- Hole: PI and declaration actions
Need to assign separate actions to PI's and declarations by name.
As a desirable side effect, this will allow loading tagsets and
interpreting DTD's. Strictly speaking the name is the name of an
application, not of an operation. PI's may have to be handled even
when passively copying!
(1998-07-15)
- Bug: Node construction must set handler.
All new nodes have to have their handler set properly; this means
replacing almost all new ParseTreeWhatever with
createWhatever operations on the current
Tagset. Easily enforced by eliminating constructors without handlers.
(1998-07-10)
- Bug: ProcessedContent doesn't insert controls.
Neither does its predecessor, InterFormContent.
(1998-07-08)
- Impr: Parser should put text in
next
(instead of putting it in nextText and increasing the
number of tests per token).
(1998-06-29)
- Hole: Ought to have
NamedChildList
for the children of things like EntityTable and
Tagset.
(1998-06-29)
-- Actually it might be best to use a ParseTreeTable as a wrapper.
- Impr: Upgrade
handle methods.
The static handle(element) methods in handler
subclasses should pick up a static instance if possible, rather than
just calling new. (See also ''Impr'' of 1998-06-15.)
(1998-06-25)
- Impr: Tagset should extend EntityTable, or reference one.
This will make it easier to define entities and tags in the same DTD.
(1998-06-19)
- Impr: PUT_VALUE action code needed
Get a value for the node and put that. Needs a corresponding
putValue(node, context, output) in Action; the
default should be to copy the node. Could be used for entity expansion
in attributes.
(1998-06-18)
- Impr: Caching for subProcess stack
ContextStack should have a free link to a free
sub-processor, to be used instead of allocating one.
(1998-06-18)
-- subProcess clears free; caller does
p.cleanup() when finished to restore it.
- Impr:
NamedActiveNode values
Named nodes should have defined, assigned and default values. Only
defined value should be the children; it would be used for the
definition of a name. Want to be able to share assigned
values; the NodeList of an assigned value would be a copy (just like we
presently do for attribute lists, and for the same reason).
- Impr:
ParseTreeElement hasActiveAttributes
ParseTreeElement should keep information about whether any of its
attributes contain entities that need to be expanded. Should be much
faster, since most don't and it's known at parse time.
The other possibility is lazy evaluation -- don't even look at them
until they're needed when converting an attribute's value to a string.