A command-line tool for combining and cleaning large word list files.
A throw of the dice will never abolish chance. — Stéphane Mallarmé
Tidy aims to help users create "better" word lists -- generally word lists that will be used to create passphrases like "block insoluble cardinal discounts".
Tidy performs basic list-cleaning operations like removing duplicate words and blank lines by default. It additionally provides various optional standardizations and filters, like lowercasing all words (-l), or removing words in with integers in them (-I), as well as protections against rare-but-possible passphrase pitfalls, such as prefix codes (-P) and low minimum word lengths (see below for explanations).
Tidy also can make word lists more "typo-resistant" by enforcing a minimum edit distance (-d), removing homophones and/or enforcing a unique prefix length (-x), which can allow users to auto-complete words after a specified number of characters.
Tidy can be used to create new word lists (for example, if given more than one list, Tidy will combine and de-duplicate them) with desirable qualities. You can obviously edit existing word lists.
Given a text file with one word per line, this tool will create a new word list in which...
and print that new word list to the terminal or to a text file.
Optionally, the tool can...
-l)-d/-D)-P)-S)-q)-x)-z)--dice)-A, -AA, -AAA, or -AAAA depending on how much information you want to printed)and more!
NOTE: If you do NOT want Tidy to sort list alphabetically, you can use the --no-sort option.
Usage: tidy [OPTIONS] <Inputted Word Lists>...
Arguments:
<Inputted Word Lists>...
Word list input files. Can be more than one, in which case they'll be
combined and de-duplicated. Requires at least one file
Options:
-a, --approve <APPROVED_LIST>
Path(s) for optional list of approved words. Can accept multiple files
-A, --attributes...
Print attributes about new list to terminal. Can be used more than once to
print more attributes. Some attributes may take a nontrivial amount of time
to calculate
-j, --json
Print attributes and word samples in JSON format
--cards
Print playing card abbreviation next to each word. Strongly recommend only
using on lists with lengths that are powers of 26 (26^1, 26^2, 26^3, etc.)
--debug
Debug mode
-d, --delete-after <DELETE_AFTER_DELIMITER>
Delete all characters after the first instance of the specified delimiter
until the end of line (including the delimiter). Delimiter must be a single
character (e.g., ','). Use 't' for tab and 's' for space. May not be used
together with -g or -G options
-D, --delete-before <DELETE_BEFORE_DELIMITER>
Delete all characters before and including the first instance of the specified
delimiter. Delimiter must be a single character (e.g., ','). Use 't' for tab
and 's' for space. May not be used together with -g or -G options
-i, --delete-integers
Delete all integers from all words on new list
-n, --delete-nonalphanumeric
Delete all non-alphanumeric characters from all words on new list. Characters
with diacritics will remain
--dice <DICE_SIDES>
Print dice roll before word in output. Set number of sides of dice. Must be
between 2 and 36. Use 6 for normal dice
--dry-run
Dry run. Don't write new list to file or terminal
-f, --force
Force overwrite of output file if it exists
--homophones <HOMOPHONES_LIST>
Path(s) to file(s) containing homophone pairs. There must be one pair of
homophones per line, separated by a comma (sun,son). If BOTH words are found
on a list, the SECOND word is removed. File(s) can be a CSV (with no column
headers) or TXT file(s)
-g, --ignore-after <IGNORE_AFTER_DELIMITER>
Ignore characters after the first instance of the specified delimiter until the
end of line, treating anything before the delimiter as a word. Delimiter must be
a single character (e.g., ','). Use 't' for tab and 's' for space. Helpful for
ignoring metadata like word frequencies. Works with attribute analysis and most
word removal options, but not with word modifications (like to lowercase).
May not be used together with -d, -D or -G options
-G, --ignore-before <IGNORE_BEFORE_DELIMITER>
Ignore characters before and including the first instance of the specified
delimiter, treating anything after the delimiter as a word. Delimiter must
be a single character (e.g., ','). Use 't' for tab and 's' for space. Helpful
for ignoring metadata like word frequencies. Works with attribute analysis
and most word removal options, but not with word modifications (like to lowercase).
May not be used together with -d, -D or -g options
--locale <LOCALE>
Specify a locale for words on the list. Aids with sorting. Examples: en-US,
es-ES. Defaults to system LANG. If LANG environmental variable is not set,
uses en-US
-l, --lowercase
Lowercase all words on new list
-M, --maximum-word-length <MAXIMUM_LENGTH>
Set maximum word length
-x, --shared-prefix-length <MAXIMUM_SHARED_PREFIX_LENGTH>
Set number of leading characters to get to a unique prefix, which can aid
auto-complete functionality. Setting this value to say, 4, means that knowing
the first 4 characters of any word on the generated list is enough to know
which word it is
-e, --minimum-edit-distance <MINIMUM_EDIT_DISTANCE>
Set minimum edit distance between words, which can reduce the cost of typos
when entering words
-m, --minimum-word-length <MINIMUM_LENGTH>
Set minimum word length
--sort-by-length
Sort by word length, with longest words first. First sorts words alphabetically, respecting inputted locale
-O, --no-sort
Do NOT sort outputted list alphabetically. Preserves original list order. Note
that duplicate lines and blank lines will still be removed
-z, --normalization-form <NORMALIZATION_FORM>
Normalize Unicode of all characters of all words. Accepts nfc, nfd, nfkc,
or nfkd (case insensitive)
-o, --output <OUTPUT>
Path for outputted list file. If none given, generated word list will be printed
to terminal
--sides-as-base
When printing dice roll before word in output, print dice values according to
the base selected through --dice option. Effectively this means that letters will
be used to represent numbers higher than 9. Note that this option also 0-indexes
the dice values. This setting defaults to `false`, which will 1-indexed
dice values, and use double-digit numbers when necessary (e.g. 18-03-08)
--print-first <PRINT_FIRST>
Just before printing generated list, cut list down to a set number of
words. Can accept expressions in the form of base**exponent (helpful
for generating diceware lists). Words are selected from the beginning
of processed list, and before it is sorted alphabetically
--print-rand <PRINT_RAND>
Just before printing generated list, cut list down to a set number of words.
Can accept expressions in the form of base**exponent (helpful for generating
diceware lists). Cuts are done randomly
--quiet
Do not print any extra information
-I, --remove-integers
Remove all words with integers in them from list
-N, --remove-nonalphanumeric
Remove all words with non-alphanumeric characters from new list. Words
with diacritics will remain
--remove-nonalphabetic
Remove all words with non-alphabetic characters from new list. Words with
diacritcis and other non-Latin characters will remain
-L, --remove-non-latin-alphabetic
Remove all words with any characters not in the Latin alphabet (A through
Z and a through z). All words with accented or diacritic characters
will be removed, as well as any words with puncuation and internal whitespace
-C, --remove-nonascii
Remove all words that have any non-ASCII characters from new list
-P, --remove-prefix
Remove prefix words from new list
-S, --remove-suffix
Remove suffix words from new list
-r, --reject <REJECT_LIST>
Path(s) for optional list of words to reject. Can accept multiple files
-s, --samples
Print a handful of pseudorandomly selected words from the created list
to the terminal. Should NOT be used as secure passphrases
-K, --schlinkert-prune
Use Sardinas-Patterson algorithm to remove words to make list
uniquely decodable. Experimental!
--skip-rows-start <SKIP_ROWS_START>
Skip first number of lines from inputted files. Useful for dealing
with headers like from PGP signatures
--skip-rows-end <SKIP_ROWS_END>
Skip last number of lines from inputted files. Useful for dealing
with footers like from PGP signatures
-q, --straighten
Replace “smart” quotation marks, both “double” and ‘single’, with
their "straight" versions
--take-first <TAKE_FIRST>
Only take first N words from inputted word list. If two or more word
list files are inputted, it will combine all given lists by alternating words
from the given word list files until it has N words
--take-rand <TAKE_RAND>
Only take a random N number of words from inputted word list. If two or more
word lists are inputted, it will combine arbitrarily and then take a random
N words. If you're looking to cut a list exactly to a specified size,
consider print-rand or whittle-to options
-W, --whittle-to <WHITTLE_TO>
Whittle list exactly to a specified length, only taking minimum number
of words from the beginning of inputted list(s). If the outputted list
is not exactly the specified length, it will try again by taking a
different amount of words form input list(s). As a result, this using this
option may cause Tidy to take a moment to produce the finished list. Can
accept expressions in the form of base**exponent (helpful for generating
diceware lists).
This option should generally only be used if all of the following conditions
are met: (a) the inputted word list is sorted by desirability (e.g. ordered
by word frequency); (b) the user is either removing prefix words, removing
suffix words, or doing a Schlinkert prune; (c) the user needs the resulting
list to be a specified length.
Optionally can also take a "starting point" after a comma. For example,
--whittle-to 7776,15000 would start by taking the first 15,000 words from
the inputted list(s) as a first attempt at making a list of 7,776 words,
iterating if necessary.
-h, --help
Print help (see a summary with '-h')
-V, --version
Print versiontidy --output new_list.txt word_list1.txt word_list2.txt Combines the word lists in word_list1.txt and word_list2.txt, removing whitespace, empty lines, and duplicate words into one list. It sorts this list alphabetically, and then prints this new, combined list to the specified output location, in this case: new_list.txt.
tidy -l -o new_list.txt inputted_word_list.txt Deletes whitespace, removes empty lines and duplicate words from inputted_word_list.txt. Due to the -l flag, it makes all the words lowercase. It sorts this list alphabetically and removes duplicates once again. It then prints this new list to the specified output location, in this case: new_list.txt.
tidy -l inputted_word_list.txt > new_list.txt Alternatively, you can use > to print tidy's output to a file.
tidy -lP -o new_list.txt inputted_word_list.txt Same as above, but the added -P flag removes prefix words from the list. See below for more on prefix words.
tidy -lPi -o new_list.txt inputted_word_list.txt Same as above, but the added -i flag deletes any integers in words. Words with integers in them are not removed, only the integers within them. For example, "11326 agency" becomes "agency".
tidy -lPiO -o new_list.txt inputted_word_list.txt Same as above, but the added -O flag preserves the original order of the list, rather than sort it alphabetically. Note that duplicates and blank lines are still removed.
tidy -I -o new_list.txt inputted_word_list.txt Using the -I flag removes any words with integers from the list. For example, "hello1" would be completely removed from the list, since it has an integer in it. Note that this is distinct from the lowercase -i flag, which would leave the word "hello" on the resulting list (removing the "1").
tidy -AA -I -o new_list.txt inputted_word_list.txt Adding -AA prints some information about the created list to the terminal. You can add up to 4 A flags to get the maximum amount of information that Tidy can print about a list. See below for more information.
tidy -l -o new_list.txt -r profane_words.txt inputted_word_list.txt Similar to above, but ensures that none of the words in the profane_words.txt file make it on to the final list that is printed to new_list.txt. The reject list is case sensitive, so you may want to run it through tidy using the -l flag before using it. (You can find lists of profane words here and here.)
tidy -l -o new_list.txt -a approved_words.txt inputted_word_list.txt Similar to above, but ensures that only words in the approved_words.txt file make it on to the final list that is printed to new_list.txt. The approved list is case sensitive. (On Mac and some Linux distributions, /usr/share/dict/words should contain a list of words for spellcheck purposes.)
tidy -l -o new_list.txt --homophones homophone_pairs.txt inputted_word_list.txt Similar to above, but expects homophones_pairs.txt to be a list of homophones pairs separated by a comma ("right,write" then next line: "epic,epoch"). If both words in the pair are on the inputted_word_list, Tidy will remove the second one. If only one of the words in the pair are on the list, Tidy won't remove it. Must be only two words per line.
tidy -lA -m 3 -o new-list.txt inputted_word_list.txt Similar to above, but the -m 3 means new list won't have any words under 3 characters in length. Have Tidy also print some attributes about the new list to the terminal screen.
tidy -z nfkd --locale fr -o bip-0039/french.txt --force bip-0039/french.txt Verify that the BIP-0039 French list is (a) normalized to Unicode Normalization Form Compatibility Decomposition (abbreviated as NFKD) (as per the BIP-0039 specification) and (b) sorted appropriately for the French language (thanks to specifying --locale fr). Locales can also be specified like "en-US" or "es-ES". If a locale is not specified, locale uses system LANG. If no LANG is found, uses "en-US". This locale setting only really affects how the words on the outputted list are sorted, so it's not crucial for most use-cases to specify one.
tidy -d t -o just_the_words.txt diceware_list.txt If you've got a diceware list with numbers and a tab before each word, the -d t flag will delete everything up to and including the first tab in each line ("11133 abruptly" becomes "abruptly").
tidy --dice 6 -o diceware_list.txt just_words.txt Add corresponding dice roll numbers to a list with --dice. Can accept dice sides between 2 and 36. Each dice roll and word are separated by a tab.
tidy -P -x 4 --print-rand 7776 --dice 6 --output diceware.txt 1password-2021.txt Make a 7,776-word list from a 1Password (~18k) word list, removing prefix words and guaranteeing 4 characters can auto-complete any word. Lastly, add corresponding 6-sided dice role for each word.
tidy -o d-and-d.txt --dice 20 --print-rand 20**3 wordlist.txt Create an 8,000-word list where each word corresponds to 3 rolls of a 20-sided die (06-07-07 dragon). --print-rand randomly truncates the resulting list to the specified amount -- can accept integers (8000) or informal exponent notation (20**3).
tidy -d s --whittle-to 7776 -PlL -m 3 -M 12 --dice 6 -o wiki-diceware.txt ~/Downloads/enwiki-20190320-words-frequency-sorted.txt Carefully make a 7,776-word list by only taking the words needed from the top of ~/Downloads/enwiki-20190320-words-frequency-sorted.txt file. Assumes this file is sorted by word frequencies, with a frequency count after the word, separated by a space (example line: located 1039008). Since we only want to use the most common words, we'll use Tidy's --whittle-to option to only take exactly how many words we need to construct a list of 7,776 words. Note that this may take longer that usual Tidy executions, since Tidy will very likely need to make multiple attempts to make a list that's exactly the requested length. More info on whittle.
cargo install --git https://github.com/sts10/tidy --locked --branch main (Run this same command to upgrade Tidy.)You should then be able to run tidy --help for help text.
Uninstall Tidy by running cargo uninstall tidy.
Check the GitHub Releases page for binaries suitable for Mac, Windows, and Linux users.
To install the executable on a Linux/macOS machine, download the tidy executable and move it to somewhere in your $PATH, like $HOME/.local/bin (you can do this on the command line with something like mv ~/Downloads/tidy ~/.local/bin/).
cargo testcargo doc --document-private-items --no-deps. Add --open flag to open docs after generation. Locally, docs are printed to ./target/doc/tidy/index.html.cargo deny check licenses (requires that you have cargo-deny installed locally)Pull Requests welcome!
Note when using Tidy to audit a list: Tidy will remove blank lines and duplicate lines (words) before calculating these list attributes. For example, if you're 4,000-word list has, say, 5 duplicate words, Tidy will report that the list has 3,995 words. No warning of duplicate words is given.
If you really want to audit a word list, without making changes to it, try Word List Auditor.
That said, Tidy can calculate different attributes about a created list. tidy -AAAA -G t --dry-run eff_long_list.txt prints:
Attributes of new list
----------------------
List length : 7776 words
Mean word length : 6.99 characters
Length of shortest word : 3 characters (aim)
Length of longest word : 9 characters (zoologist)
Free of prefix words? : true
Free of suffix words? : false
Uniquely decodable? : true
Entropy per word : 12.925 bits
Efficiency per character : 1.849 bits
Assumed entropy per char : 4.308 bits
Above brute force line? : true
Shortest edit distance : 1
Mean edit distance : 6.858
Longest shared prefix : 8
Unique character prefix : 9Using the --samples flag will print 5 sample passphrases to the terminal. (Note that these sample passphrases should not be used for security purposes, as Tidy has not been audited.)
Word samples
------------
destruct subpar dizzy outshine stipend ovary
slapstick hastily tremor visibly gizzard unloaded
salaried unwieldy churn vanity speak vessel
deserve humble pantyhose dayroom reprise unnatural
vascular stencil visible sporty embellish submarineWhen counting the length of a word, Tidy counts the number of grapheme clusters in the word. Generally, less common characters like accented letters and emoji all count as 1 grapheme cluster and thus, to Tidy, one character. I believe this better fits with how us humans intuitively count characters in a string/word.
In general, Tidy expects inputted files to have one word per line.
Tidy supports \n and \r\n line endings.
In both Tidy's code and documentation, "remove" means that a word will be removed (e.g. words with integers will be removed from the list), while "delete" means that a word will only be modified (e.g. integers removed from words). Uppercase flags remove words, while lowercase flags delete specified characters. All delete calls and word modifications (like "to lowercase") occur before any remove call.
Tidy does its best to work well with all languages. That said, I'm an English speaker and have not tested Tidy with other languages all that much.
There are a few steps you can take to help Tidy produce a good word list in all languages.
If you're using Tidy to work a word list with accented characters, it is highly recommended that you:
-z nfc or -z nfkd). This will better ensure that there are no duplicate-looking words on the list, which could cause Tidy and others to over-estimate the strength of passphrases generated from the outputted list. Note that if you're passing a reject list file or approved list file to Tidy, you should normalize those lists before using them. For example: tidy -z nfc --locale ES-es -l --force -o profane-spanish-words.txt profane-spanish-words.txt && tidy -z nfc --locale ES-es -r profane-spanish-words.txt -o my-new-spanish-word-list.txt -l a-bunch-of-spanish-words.txt--locale fr or --locale ES-es). This will ensure that the outputted list is sorted correctly.-q or --straighten option to standardize these characters across all words on the new list.See this blog post for more. If you find Tidy not performing as expected with non-English words, please open an Issue on this repository with an example.
If passphrases from your list will ever be spoken out loud, you may want to consider removing homophones -- words that sound alike -- from your list.
I'd say that Tidy offers two ways of dealing with homophones.
Given a pair of homophones, like "sun" and "son":
tidy with a flag like --homophones ../homophones/homophone-lists/homophones-large-as-pairs.txt (link). This will let either "sun" or "son" on your list but NOT both.-r ../homophones/homophone-lists/cleaned-as-singles.txt (link). This will remove both "sun" and "son" from your generated list before its outputted.If you're looking for a relatively long list of English homophones, I'd humbly point you to this other project of mine.
If a word list is "uniquely decodable" that means that words from the list can be safely combined without a delimiter between each word, e.g. enticingneurosistriflecubeshiningdupe.
As a brief example, if a list has "boy", "hood", and "boyhood" on it, users who specified they wanted two words worth of randomness (entropy) might end up with "boyhood", which an attacker guessing single words would try. Removing the word "boy", which makes the remaining list uniquely decodable, prevents this possibility from occurring.
To make a list uniquely decodable, Tidy removes words. Tidy offers three (3) distinct procedures to make cuts until a list is uniquely decodable. Users can (1) remove all prefix words, (2) remove all suffix words, or (3) perform "Schlinkert pruning," a procedure based on the Sardinas–Patterson algorithm that I developed for Tidy. Note that Schlinkert pruning a long inputted word list may take hours or days; removing prefix or suffix words should be significantly quicker. You can learn more about uniquely decodable codes and Schlinkert pruning by reading this blog post.
Tidy can also simply check if the inputted list is (already) uniquely decodable. It does this using the Sardinas–Patterson algorithm. You can do this by passing Tidy four attributes flag (-AAAA).
Tidy offers an option --whittle-to. This option should only be used in specific situations -- users generally should prefer --print-rand or --print-first options. The situation where whittling gives an advantage over the print options is when the following conditions are met:
(a) the inputted word list is sorted by desirability (e.g. ordered by word frequency) and
(b) the user is either removing prefix words (-P), removing suffix words (-S), and/or doing a Schlinkert prune (-K).
To see why whittling is best for this particular situation, see this document.
Tidy allows users to set a maximum shared prefix length.
Setting this value to say, 4, means that knowing the first 4 characters of any word on the generated list is sufficient to know which word it is.
On this example generated list where we told Tidy to make the maximum shared prefix length 4 characters, we'd know that if a word starts with "radi", we know it must be the word "radius" (if "radical" had been on the list, Tidy would have removed it).
This is useful if you intend the list to be used by software that uses auto-complete. For example, a user will only have to type the first 4 characters of any word before a program could successfully auto-complete the entire word.
(Note that this setting is distinct from the operation of eliminating prefix words, though can be used in conjunction with that feature.)
Use the attributes flag twice (-AA) to get information about shared prefix length for a generated list. Tidy will print both "Longest shared prefix" and "Unique character prefix" (which is longest shared prefix + 1).
If we take the entropy per word from a list (log2(list_length)) and divide it by the average word length of words on the list, we get a value we might call "efficiency per character". This just means that, on average, you get E bits per character typed.
If we take the entropy per word from a list (log2(list_length)) and divide it by the length of the shortest word on the list, we get a value we might call "assumed entropy per char" (or character).
For example, if we're looking at the EFF long list, we see that it is 7,776-words long, so we'd assume an entropy of log27776 or 12.925 bits per word. The average word length is 7.0, so the efficiency is 1.8 bits per character. (I got this definition of "efficiency" from an EFF blog post about their list.) And lastly, the shortest word on the list is three letters long, so we'd divide 12.925 by 3 and get an "assumed entropy per character" of about 4.31 bits per character.
I contend that this "assumed entropy per character" value in particular may be useful when we ask the more theoretical question of "how short should the shortest word on a good word list should be?" There may be an established method for determining what this minimum word length should be, but if there is I don't know about it yet! Here's the math I've worked out on my own.
Assuming the list is comprised of 26 unique characters, if the shortest word on a word list is shorter than log26(list_length), there's a possibility that a user generates a passphrase such that the formula of entropy_per_word = log2(list_length) will overestimate the entropy per word. This is because a brute-force character attack would have fewer guesses to run through than the number of guesses we'd assume given the word list we used to create the passphrase.
As an example, let's say we had a 10,000-word list that contained the one-character word "a" on it. Given that it's 10,000 words, we'd expect each word to add an additional ~13.28 bits of entropy. That would mean a three-word passphrase would give users 39.86 bits of entropy. However! If a user happened to get "a-a-a" as their passphrase, a brute force method shows that entropy to be only 14.10 bits (4.7 * 3 words). Thus we can say that it falls below the "brute force line", a phrase I made up.
To see if a given generated list falls above or below this line, use the -A/--attributes flag.
Formula:
Where S is the length of the shortest word on the list, 26 is the number of letters in the English alphabet, and M is max list length: M = 2S * log2(26). Conveniently, this simplifies rather nicely to M = 26S.
(or in Python: max_word_list_length = 26**shortest_word_length)
| shortest word length | max list length |
|---|---|
| 2 | 676 |
| 3 | 17576 |
| 4 | 456976 |
| 5 | 11881376 |
If we go by a 1951 Claude Shannon paper, each letter in English actually only gives 2.6 bits of entropy. Users can see if their generated word list falls above this (stricter) line -- which I've dubbed the "Shannon line" -- by using the -A/--attributes flag.
Formula:
Where S is the length of the shortest word on the list and M is max list length: 2S * 2.6 = M
(or in Python: max_word_list_length = 2**(shortest_word_length*2.6), which, to preserve correct number of significant digits, should be max_word_list_length = 6.1**shortest_word_length)
| shortest word length | max list length |
|---|---|
| 2 | 37 |
| 3 | 226 |
| 4 | 1384 |
| 5 | 8445 |
| 6 | 51520 |
As you can see, the Shannon line is quite a bit more "strict" than the brute force line.
A word list of 7,776 words "fits" nicely into 5 6-sided dice rolls. But not all word lists are 7,776 words long.
If you'd like some help figuring out how to fit your list to a number of dice rolls, another tool I wrote called Dice Tailor might help.
Tidy's function for removing characters on either side of a given delimiter uses a library called memchr, which "provides heavily optimized routines for searching bytes." The optimization gained from using this crate is far from noticeable or necessary for most uses of Tidy -- using Rust's built-in find is not much slower -- but I figured the extra speed was worth the dependency in this case.
See this repo for more information.
This project uses cargo-dist to create releases.
Some of my personal docs are here; but basically, cargo install cargo-dist. When you're ready to cut a new release, test the current state of the project with cargo dist build and cargo dist plan. If that went well, create a new git tag that matches the current project version in Cargo.toml with git tag vX.X.X. Finally, run git push --tags to kick off the release process. GitHub will handle it from here -- check your project's GitHub Releases page in about 5 to 10 minutes.
tidy --no-sort for de-duplicating large word lists. Written in C.