factor/TODO.FACTOR.txt

104 lines
2.9 KiB
Plaintext

- quit responder breaks with multithreading
- nicer way to combine two paths
- don't show listener on certain commands
- plugin should not exit jEdit on fatal errors
- wordpreview: don't show for string literals and comments
- alist -vs- assoc terminology
- NPE in activate()/deactivate()
- write-icon kind of messy; " " should be output by the listener
- f usages. --> don't print all words
- file responder: don't show full path in title
- clean up listener's action popups
- jedit ==> jedit-word, jedit takes a file name
- introduce ifte* and ?str-head/?str-tail where appropriate
- namespace clone drops static var bindings
- when running (inf, .factor-rc not loaded
+ bignums:
- move some s48_ functions into bignum.c
- remove unused functions
- add a socket timeout
- >lower, >upper for strings
- accept multi-line input in listener
+ docs:
- examples of assoc usage
- unparse examples, and difference from prettyprint
- review doc formatting with latex2html
- recursion -vs- iteration in vectors chapter, and combinator
construction
- [, , ,] -- mention that , are usually in nested words
- finish namespaces docs
- mention word accessors/mutators
- to document:
>r r> (earlier on?)
continuations
streams
multitasking
unit testing
+ tests:
- java factor: equal numbers have non-equal hashcodes!
- FactorLib.equal() not very good
- investigate mandel.factor
+ listener/plugin:
- NPE in ErrorHighlight
- some way to not have previous definitions from a source file
clutter the namespace
- use inferior.factor for everything not just listener
- fedit broken with listener
- maple-like: press enter at old commands to evaluate there
- input style after clicking link
- listener backspace overzealous
- completion in the listener
- special completion for USE:/IN:
- inspector links when describe called without object path
+ native:
- fix error postoning -- not all errors thrown by i/o code are
postponed
- sbuf-hashcode
- vector-hashcode
- irc: stack underflow?
- gc call in the middle of some ops might affect callstack
- better i/o scheduler
+ JVM compiler:
- compiled stack traces broken
- save classes to disk
- tail call optimization broken again
- don't compile inline words
- recursive words with code after ifte
- less unnecessary args to auxiliary methods
- inlining tail-recursive immediates
- direct stack access leaks memory on stack
- unnecessary local allocation: max is instance var, but several methods
get compiled.
+ misc:
- some way to run httpd from command line
- don't rehash strings on every startup
- 'cascading' styles
- ditch expand
- set-object-path
+ httpd:
- 'default responder' for when we go to root
- wiki responder:
- port to native
- text styles
- log with date
basic authentication, using httpdAuth function from a config file
- file responder; last-modified field