philhanna / crossword Goto Github PK

The user should be able to start typing in the "Word" input box.

Clear all the letters that are not part of an already completed word

Need some externally visible web server that supports WSGI.
Possibilities:

• AWS
• Apache on my own development machine

Once a puzzle is complete, need a way to export it to PDF in various formats. To begin with, I'm only using the NYTimes format. See Times submissions for details.
I have working code in nytimes_output.py, but it requires a file name for the output and a puzzle object. The GUI needs to supply these two things.

Right now, I have it as a top-level menu option, but maybe it should be part of the puzzle menu?

The modal dialog edit-word obscures the puzzle.

The new grid dialog prompts for an odd integer for the grid size. The input field is validated as being:

• not empty
• numeric
• an integer
• odd

However, the program does not check whether it is positive.

Need to add this capability

Currently the system uses gridnxn or puzzlenxn as the initial name when a new grid or puzzle is created. Better to switch to (untitled). Then do not allow file to be saved when untitled, only "Saved As" a real name.

Steps to reproduce:

1. Open a puzzle
2. Click on a word start
3. In the edit word dialog, shorten the displayed word to just one character
4. Click OK
   Error occurs:
   File "crossword/webapp.py", line 256, in edit_word_screen
   word.set_text(text)
   File "crossword/word.py", line 46, in set_text
   letter = text[i]
   IndexError: string index out of range

In publish -> New York Times format, the output is generated and downloaded, but the puzzle chooser dialog is still present. Need to redirect to main screen.

See closed issue #46

Steps to reproduce

1. Open a grid
2. Click "Show grid statistics" in menu
3. Close statistics window
4. Click "Show grid statistics" again

The size, errors, word count, and word length table are repeated, concatenated with the old.

At present, you have to be completely done with designing a grid (the black cells) before you can create a puzzle from it. This makes it inconvenient if you have been working on a puzzle for a while and then realize that you need to change the grid structure. You don't want to lose all the work you've done.

Proposed change

• Add a menu option under Puzzle to extract and save this puzzle's grid
• Add a menu option under Puzzle to use a different grid
• Can only be done if the grids are the same size
• The new grid is merged into the existing puzzle:
  • The puzzle's current black cells are set to blanks
  • The new grid's black cells are copied to the puzzle

This consists of several tasks:

• Change the nytimes_output.py to use DOM rather than generating HTML manually
• Do not automatically write the HTML and SVG output to files - return as strings
• In webapp.publish_nytimes_screen(), create a .zip file with these two files
• Return download link

The bottom border is too small. Same is true with puzzle chooser dialog.

This seems to be left over. I now put the word length in the header, but I still see the tool tip when I hover over the label.

Menu option to see:

• Validity
• Word count
• Words by length

Currently this information is only seen when the grid is saved.

Sometimes when editing a grid, the cells that should be numbered don't always get new numbers.
I have noticed this when I click a cell and then click it again to clear it.
I have also noticed that when I click outside the grid it seems to cause the renumbering.

Need to be able to export a puzzle to .csv and import from .csv. This will allow someone to work on the clues outside of the application. The system already has the capability to do this (see clue_export_visitor), it just needs to be integrated with the menu.

The raw .csv file can be edited directly with Microsoft Excel, or uploaded to Google Drive so that it can be shared with whoever needs to update it.

Currently, the user can click on any set of cells in the grid to make them black cells.
However, not all these configurations are valid. For example, the New York Times
requires that:

• Crosswords must have black square symmetry, which typically comes in the form of 180-degree rotational symmetry;
• Crosswords must have all-over interlock;
• Crosswords must not have unchecked squares
 (i.e., all letters must be found in both Across and Down answers);
• All answers must be at least 3 letters long;
• Black squares should be used in moderation.
  (Source: NYTimes submissions guidelines)

When the user chooses "Save grid" or "Save grid as...", the system should ensure that these guidelines are met. If not, the save should be prevented and message should be flashed.

Unit tests run fine in PyCharm, but not from a command line. test_wordlist fails because it can't find ../words.

From the beginning of the editing session for a puzzle, keep track of each time the Edit Word OK button has been clicked, and implement undo/redo for the word value.

Currently, the validation messages are all run together, separated by titles that start and end with "***".
Need to return the three validation sets individually, so that they can be displayed better in the UI.

It's clumsy to have to dismiss a message dialog. Better to have a place in the main screen where messages can be displayed. There is frequently only one message; perhaps a w3-panel at the bottom of the screen? Still with a close X.

grid-chooser-ajax.js currently builds anchor tags within list items manually with
string concatenation. This is fragile and error-prone. Better to build DOM child elements
with the DOM API, as in do-suggest-word.js.

    links = "";
    for (i = 0; i < grid_list.length; i++) {
        filename = grid_list[i]
        href = build_url(filename)
        quoted_href = "'" + href + "'"
        anchor = "<a href=" + quoted_href + ">" + filename + "</a>"
        listitem = "<li>" + anchor + "</li>"
        links += listitem + "\n"
    }
    document.getElementById("grid-list").innerHTML = links

Getting a key error in puzzle_screen() line 283:
puzzle = Puzzle.from_json(session['puzzle'])

In Firefox, if I click on a grid, the parent x and y coordinates of the containing element are different than they are in Chrome. This means that I am not calculating the row and column of the click properly. This affects grid.js and puzzle.js.

Steps to reproduce

1. Bring up sample grid-1 in Firefox
2. Click 22 across to create a black cell
3. Click 22 across to remove this black cell

The 22 across black cell remains, and (1,1) is now black

Progress so far

There are two Javascript functions that use the click event:

• grid_click(event), which is triggered when the user clicks on a cell in the grid screen
• getRC(event), which is triggered when the user clicks on a cell in the puzzle screen

Both of them use code like this:

 var x = event.offsetX;
 var y = event.offsetY;

Apparently the Firefox event model is different. I fiddled with it for several hours and couldn't make any headway.

Be able to print the puzzle in final published form, without the answers, so that I can just hand it to someone to solve.

All the methods should be static.

ME: Delete myfile
COMPUTER: Are you sure (Y/N)?
ME: Yes
COMPUTER: Are you really sure (Y/N)?
ME: Yes
COMPUTER: Are you really really sure (Y/N)?
ME: Yes
COMPUTER: (deletes myfile)
ME: Whoops, damn!

The problem is that I meant to delete myotherfile.

"Are you sure" messages shouldn't just ask you to confirm the action you want to perform, but the thing you want to perform the action on. The message should include the name of the thing:

COMPUTER: Are you sure you want to delete "myfile" (Y/N)?
ME: No
ME: Phew!

If a "suggest word" dropdown has been populated, it needs to be cleared before the edit word dialog is displayed.

User can enter a regular expression in the word, like [AEIOU] for a vowel, or [^AEIOU] for a consonant. This is extremely useful. But if user clicks OK while the regex is still in the input box, it overlays the text of the word.

I think it's good enough to just check for illegal characters in Javascript and open an alert if there are any.

If a grid or puzzle has been changed since the last save (or save as), the system should pop up a dialog box that makes the user confirm that they want to close without saving.

I have been careless about not using var when declaring variables in javascript.

Sometimes fails to return a selection:

• If there is only one entry in the list

The README.md file should provide just the following:

• Overview
• Installation instructions

The rest of README.md should be migrated to the github wiki.

Right now, I generate HTML that links to an SVG image. I have to post-process it by opening it in Microsoft Word, fiddling around with the positioning, then saving as PDF.

It would be much more straightforward to generate the PDF directly.

This would change:

• Publish in New York Times format

Steps to reproduce:

Save

1. Open a puzzle
2. Save puzzle
3. Close puzzle

Save As

1. Open a puzzle (e.g., 'puzzle1')
2. Save as 'puzzle2'
3. Close puzzle

System pops up a close without saving dialog

Steps to reproduce

1. Open a grid
2. Save it
3. Menu does not enable "Show grid statistics"

Steps to reproduce

1. Open a puzzle
2. Double click on a numbered down cell
   Nothing happens - only across cells are working

Steps to reproduce

1. From menu, click "Publish -> New York Times format"
2. Select a puzzle

Output is generated, but no messages are seen.

In the edit word dialog, there is a suggest button. Need to hook this up with a search of the word list for matches for the word, including its actual letters and the dots that represent missing letters.

Need a shorter name for ids. Maybe two-letter abbreviation for the dialog itself, then one letter suffix for individual elements.