WHAT's Wrong (15-Jul-15)

WHAT, Wolfberg's Helpful Anagramming Tool, is essentially complete except for documentation within the program, i.e. on-line help. Also, the user guide deserves some updating, and this will be done as time permits.

Contents

• Candidates for Version 1
• Recently Reported Problems
• Known Bugs and Questionable Features
• Suggestions for Enhancements

Candidates for Version 1

These features were candidates for inclusion in Version 1, but they are missing for now. Users who would like to see these should please speak up:

• Menu and dialog support for bringing back commands from the command history is currently missing.
  
  
• 5 x 5 Grid Game cube contents could be user-specifiable, such as in the what.ini file. These are currently built-in.
  
  
• When a user gets ready to exit or to switch between lexicons of different languages, perhaps WHAT, should WHAT should produce a warning when there are wordlists that have changed and thus deserve to be exported first.
  
  
• We may support the saving and restoring of all wordlists to/from files using a single command. At present you may export or import individual wordlists one at a time.
  
  
• Provide for a translation of a command into English, so if you present a command, you can be told what it will do. This may be useful in certain situations, since the command language is intentionally terse. We will be curious how many of you have a need for this feature, espcially once you are a seasoned user.
  
  
• Optionally provide a GUI in French and/or Spanish for French and/or Spanish users. At present only the lexicons have language-specific words.
  
  

Recently Reported Problems

This section lists recently reported problems which we agree are sufficiently serious bugs to be fixed relatively soon. This section remains here, even when there are no such known bugs. These are bugs that we expect to repair soon:

• We are pleased to announce, this category has no entries today. This is likely to change as soon as many new users begin to try untested operations in the early days of WHAT's availability.

Known Bugs and Questionable Features

• Noticed mid-July, 2015 are problems when reading a .gcg file - the history is selecting the game start, but that is not what is shown on the board and with the rack. What is shown is the state after one move has been made. We expect to get this bug fixed and make Release 94, as soon as this is corrected. The problem was noticed while preparing the document on using WHAT with Scrabble® games.
• Another bug in this area of the program is when selecting the activity of playing moves during the showing of a recorded game, the button for requesting all move choices disappears. We also expect to get this bug fixed and make Release 94, as soon as this is corrected. The problem was noticed while preparing the document on using WHAT with Scrabble® games.
• You can press the Esc key to stop WHAT's performing a query. There may be some queries which cause the program to get into some processing which is uninterruptable using Esc. Unfortunately, we are not sure if this old deficiency has been fixed everywhere. This low-priority bug deserves to be fixed. Please let Mike know if you discover such a problem.
  
  
• The dialog to add rack letters and blanks should be enhanced to also allow for many kinds of sets. Until this is implemented, users must include sets in racks by typing, rather than by GUI use. This is not much of a burden, since it merely requires a comma to indicate what follows is in the rack.
  
  
• Although the WHAT command language allows for sets to be specified with letter ranges,such as  (A-E), the facilities in the "Add a Query Constituent" dialog do not include this possibility.
  
  
• Support for long command lines and pasting of long lines and multiple lines into the workspace is incomplete. These problems will be addressed eventually, but we consider these of low priority.
  
  
• There is currently a restriction that changing the workspace font size must be done right away after starting WHAT. Perhaps this restriction can be lifted, but we will pursue this at a low priority.
  
  
• If the slate contains all 3-letter words, and you perform the query 0[] (where the first character is the digit zero), the results are unexpected. They include words of length 2-4, and it is unclear what is being computed. For example, one answer is AGEE. So a repetition count of 0 for a word reference seems to not work; it is not caught as an error today.
  
  
• The query *coil/~p produces only 5-letter words, even though words that end in coil are of lengths up to 9 letters. I was expecting longer words in the answer. Perhaps the use of an asterisk in a blank circulation query represents one blank and not an arbitrary number. If this is true, something should be so mentioned in the user guide.
  
  
• The relationship between steal queries and word length filtering should be studied and either clarified or repaired. There may be a problem in this area. If you ask for steals of WOLFBERG, you get 3 11-letter words. If you seek 7-letter words in WOLFBERG, you are told there are none, but if you combine the requests, you are shown 28 words that have 6 letters from WOLFBERG and one additional letter. This is surprising, but is it a bug or a feature?
  
  
• There are some known GUI problems:
  • When dealing with a Spansh lexicon, the tilde-N character may be typed in by typing CTRL-N. There is a subtle bug which seems to be unfixable due to the available facitilies of the underlying system upon WHAT is built, and this is whether a tilde-N is uppercase or lowercase may be wrong when it is the very first character typed into WHAT and when the Caps Lock is on when WHAT is started. This seems to be of almost no importance to be considered, but since it is really a bug, we are noting it here.
    
    
  • Pressing Ctrl+E in the workspace seems to indent the display of the current line to the middle of the workspace. This is not intentional, and we have not understood on what basis this happens, but it seems to be harmless.
    
    
  • Pressing Ctrl+R in the workspace seems to indent the display of the current line to the right of the workspace. This is not intentional, and we have not understood on what basis this happens, but it seems to be harmless.
    
    
  • Sometimes the green characters that indicate blanks get messed up. If you suspect the display is incorrect with respect to these green characters, request the presentation again, and it may very well be done correctly. This problem seems to stem from timing problem with the way WHAT uses the underlying system upon which it is built, and thus far fixing this has eluded us.
    
    
  • When the cursor is in the Definitions window, pressing the right mouse button brings up a menu. This feature is unintentional, but is harmless.
    
    
  • Paste a long line into WHAT that wraps. The cursor is placed on the first line of the command at a position one line above where it ought to be. WHAT is not expecting one command to span more than one line. For now, this can be avoided by not wrapping.
    
    
  • Underlined characters in the WHAT menus do not show up until the Alt key is pressed. Perhaps they should always be there, but we have not found a way.
    
    
  • The set operation commands (union, intersection, and set difference) are not specifiable in the GUI.

Suggestions for Enhancements

• With Version 2.0, WHAT understands a board configuration so that it can report highest scoring plays on a board with letters on it. Game recording and playback are supported. We have not yet confirmed that this new feature is functioning fully for Spanish lexicons.
  
  
• The commands /Z and /-Z are meant to control whether blanks score zero or the value of the reprented tile. Setting this option has no effect at this point. We will look into how to fix this.
  
  
• In flashcarding, where a card goes back into the pack could be controlled by something on the card.
  
  
• Some folks would like a way to input the answers for a flashcard to then be checked. Only some of this is now supported.
  
  
• Generalize the feature of presenting unhooks to allow for more than letter to be unhooked, and provide a filter for whether a word being referenced as a word reference has such an unhook.
  
  
• Support a scripting language for making extensions.
  
  
• Type in some letters, such as a rack of 7 letters, and then be shown a list of all the interesting things you may want to ask about them.
  
  
• Some higher-level dialogs can be built on top of the existing program to do standard simple-to-state tasks, along the lines of what is supported by the LeXpert program.
  
  
• WHAT shows what a blank can be in lime green, but with sets there are various blank-like query characters. For example, a letter matched to a vowel is shown in green. Should it be any color other than for a normal character? Should something be green whenever it is not coming from a singleton set?
  
  
• Support the highlighting in some way (italic, bold, etc.) for certain characters, such as those which are green in the workspace or those which are hook letters.
  
  
• Should there be an end-of-word set constituent? Underscore may be a candidate character for this. This would make some queries simpler, such as "what words have a Q that is not followed by a U", since the separate case has to be covered for when the word ends in a Q.
  
  
• We first decided on /F for commands involved with files. We have also considered using /X and /M for export and import. We may yet want to make this change, after thinking about it some more. Appending would probably then be specified with /+X.
  
  
• LeXpert can find words that don't match a condition. Is this idea worthy of support in WHAT? Does the concept go along with the WHAT design?
  
  
• We may want to provide facilities in WHAT to produce likelihoods of certain rack leaves leading to words of a certain length, especially 7- and 8-letter words. For example, you could compute the likelihood of bingoing with a leave of 2-4 of the letters in ERST.
  
  
• If the speed of anagramming is found to be a bit slow, a secondary word tree per lexicon with words in canonical order may prove to be worthwhile. This may become apparent if the above idea of computing likelihoods is done using simulation instead of computation.
  
  
• We may support the outputting to a file set definitions by themselves, or this may be part of a more general feature of outputting commands that represent the state of WHAT, such that they can be used in a future WHAT session to get back to the current state. This would not include the contents of the slate and other wordlists.
  
  
• When solving 5 x 5 Grid Games, show on the grid where a particular word is, such as by highlighting.
  
  
• In 5 x 5 Grid Game solving, allow blanks and sets for any cube. This may turn out to be easy, but does anyone really want such support?
  
  
• When one makes a query resulting in a large number of words to present, a Large Output Dialog comes up, and it does not offer an option to see the tail end of the slate. Is this something users want?
  
  
• A user-defined set cannot be the argument in a filter subcommand that involves sets, such as when filtering back hooks. Is this too restrictive?
  
  
• The only sets that can be used in set operation commands (union, intersection, and set difference) are user-defined ones. This is not a serious restriction, since you can copy any set to a user-defined one.