Build an ASCII chart of the most commonly used wor

Question

Closed. This question is off-topic. It is not currently accepting answers.

Closed 6 years ago.

Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.

The challenge:

Build an ASCII chart of the most commonly used words in a given text.

The rules:

• Only accept a-z and A-Z (alphabetic characters) as part of a word.
• Ignore casing (She == she for our purpose).
• Ignore the following words (quite arbitary, I know): the, and, of, to, a, i, it, in, or, is

• Clarification: considering don't: this would be taken as 2 different 'words' in the ranges a-z and A-Z: (don and t).

• Optionally (it's too late to be formally changing the specifications now) you may choose to drop all single-letter 'words' (this could potentially make for a shortening of the ignore list too).

Parse a given text (read a file specified via command line arguments or piped in; presume us-ascii) and build us a word frequency chart with the following characteristics:

• Display the chart (also see the example below) for the 22 most common words (ordered by descending frequency).
• The bar width represents the number of occurences (frequency) of the word (proportionally). Append one space and print the word.
• Make sure these bars (plus space-word-space) always fit: bar + [space] + word + [space] should be always <= 80 characters (make sure you account for possible differing bar and word lengths: e.g.: the second most common word could be a lot longer then the first while not differing so much in frequency). Maximize bar width within these constraints and scale the bars appropriately (according to the frequencies they represent).

An example:

The text for the example can be found here (Alice's Adventures in Wonderland, by Lewis Carroll).

This specific text would yield the following chart:

 _________________________________________________________________________
|_________________________________________________________________________| she 
|_______________________________________________________________| you 
|____________________________________________________________| said 
|____________________________________________________| alice 
|______________________________________________| was 
|__________________________________________| that 
|___________________________________| as 
|_______________________________| her 
|____________________________| with 
|____________________________| at 
|___________________________| s 
|___________________________| t 
|_________________________| on 
|_________________________| all 
|______________________| this 
|______________________| for 
|______________________| had 
|_____________________| but 
|____________________| be 
|____________________| not 
|___________________| they 
|__________________| so 

For your information: these are the frequencies the above chart is built upon:

[('she', 553), ('you', 481), ('said', 462), ('alice', 403), ('was', 358), ('that
', 330), ('as', 274), ('her', 248), ('with', 227), ('at', 227), ('s', 219), ('t'
, 218), ('on', 204), ('all', 200), ('this', 181), ('for', 179), ('had', 178), ('
but', 175), ('be', 167), ('not', 166), ('they', 155), ('so', 152)]

A second example (to check if you implemented the complete spec): Replace every occurence of you in the linked Alice in Wonderland file with superlongstringstring:

 ________________________________________________________________
|________________________________________________________________| she 
|_______________________________________________________| superlongstringstring 
|_____________________________________________________| said 
|______________________________________________| alice 
|________________________________________| was 
|_____________________________________| that 
|______________________________| as 
|___________________________| her 
|_________________________| with 
|_________________________| at 
|________________________| s 
|________________________| t 
|______________________| on 
|_____________________| all 
|___________________| this 
|___________________| for 
|___________________| had 
|__________________| but 
|_________________| be 
|_________________| not 
|________________| they 
|________________| so 

The winner:

Shortest solution (by character count, per language). Have fun!

---

Edit: Table summarizing the results so far (2012-02-15) (originally added by user Nas Banov):

Language          Relaxed  Strict
=========         =======  ======
GolfScript          130     143
Perl                        185
Windows PowerShell  148     199
Mathematica                 199
Ruby                185     205
Unix Toolchain      194     228
Python              183     243
Clojure                     282
Scala                       311
Haskell                     333
Awk                         336
R                   298
Javascript          304     354
Groovy              321
Matlab                      404
C#                          422
Smalltalk           386
PHP                 450
F#                          452
TSQL                483     507

The numbers represent the length of the shortest solution in a specific language. "Strict" refers to a solution that implements the spec completely (draws |____| bars, closes the first bar on top with a ____ line, accounts for the possibility of long words with high frequency etc). "Relaxed" means some liberties were taken to shorten to solution.

Only solutions shorter then 500 characters are included. The list of languages is sorted by the length of the 'strict' solution. 'Unix Toolchain' is used to signify various solutions that use traditional *nix shell plus a mix of tools (like grep, tr, sort, uniq, head, perl, awk).

Answer 1

C# - 510 451 436 446 434 426 422 chars (minified)

Not that short, but now probably correct! Note, the previous version did not show the first line of the bars, did not scale the bars correctly, downloaded the file instead of getting it from stdin, and did not include all the required C# verbosity. You could easily shave many strokes if C# didn't need so much extra crap. Maybe Powershell could do better.

using C=System.Console;   // alias for Console
using System.Linq;  // for Split, GroupBy, Select, OrderBy, etc.

class Class // must define a class
{
    static void Main()  // must define a Main
    {
        // split into words
        var allwords = System.Text.RegularExpressions.Regex.Split(
                // convert stdin to lowercase
                C.In.ReadToEnd().ToLower(),
                // eliminate stopwords and non-letters
                @"(?:\b(?:the|and|of|to|a|i[tns]?|or)\b|\W)+")
            .GroupBy(x => x)    // group by words
            .OrderBy(x => -x.Count()) // sort descending by count
            .Take(22);   // take first 22 words

        // compute length of longest bar + word
        var lendivisor = allwords.Max(y => y.Count() / (76.0 - y.Key.Length));

        // prepare text to print
        var toPrint = allwords.Select(x=> 
            new { 
                // remember bar pseudographics (will be used in two places)
                Bar = new string('_',(int)(x.Count()/lendivisor)), 
                Word=x.Key 
            })
            .ToList();  // convert to list so we can index into it

        // print top of first bar
        C.WriteLine(" " + toPrint[0].Bar);
        toPrint.ForEach(x =>  // for each word, print its bar and the word
            C.WriteLine("|" + x.Bar + "| " + x.Word));
    }
}

422 chars with lendivisor inlined (which makes it 22 times slower) in the below form (newlines used for select spaces):

using System.Linq;using C=System.Console;class M{static void Main(){var
a=System.Text.RegularExpressions.Regex.Split(C.In.ReadToEnd().ToLower(),@"(?:\b(?:the|and|of|to|a|i[tns]?|or)\b|\W)+").GroupBy(x=>x).OrderBy(x=>-x.Count()).Take(22);var
b=a.Select(x=>new{p=new string('_',(int)(x.Count()/a.Max(y=>y.Count()/(76d-y.Key.Length)))),t=x.Key}).ToList();C.WriteLine(" "+b[0].p);b.ForEach(x=>C.WriteLine("|"+x.p+"| "+x.t));}}

Answer 2

Java - 886 865 756 744 742 744 752 742 714 680 chars

• Updates before first 742: improved regex, removed superfluous parameterized types, removed superfluous whitespace.

• Update 742 > 744 chars: fixed the fixed-length hack. It's only dependent on the 1st word, not other words (yet). Found several places to shorten the code (\\s in regex replaced by and ArrayList replaced by Vector). I'm now looking for a short way to remove the Commons IO dependency and reading from stdin.

• Update 744 > 752 chars: I removed the commons dependency. It now reads from stdin. Paste the text in stdin and hit Ctrl+Z to get result.

• Update 752 > 742 chars: I removed public and a space, made classname 1 char instead of 2 and it's now ignoring one-letter words.

• Update 742 > 714 chars: Updated as per comments of Carl: removed redundant assignment (742 > 730), replaced m.containsKey(k) by m.get(k)!=null (730 > 728), introduced substringing of line (728 > 714).

• Update 714 > 680 chars: Updated as per comments of Rotsor: improved bar size calculation to remove unnecessary casting and improved split() to remove unnecessary replaceAll().

---

import java.util.*;class F{public static void main(String[]a)throws Exception{StringBuffer b=new StringBuffer();for(int c;(c=System.in.read())>0;b.append((char)c));final Map<String,Integer>m=new HashMap();for(String w:b.toString().toLowerCase().split("(\\b(.|the|and|of|to|i[tns]|or)\\b|\\W)+"))m.put(w,m.get(w)!=null?m.get(w)+1:1);List<String>l=new Vector(m.keySet());Collections.sort(l,new Comparator(){public int compare(Object l,Object r){return m.get(r)-m.get(l);}});int c=76-l.get(0).length();String s=new String(new char[c]).replace('\0','_');System.out.println(" "+s);for(String w:l.subList(0,22))System.out.println("|"+s.substring(0,m.get(w)*c/m.get(l.get(0)))+"| "+w);}}

More readable version:

import java.util.*;
class F{
 public static void main(String[]a)throws Exception{
  StringBuffer b=new StringBuffer();for(int c;(c=System.in.read())>0;b.append((char)c));
  final Map<String,Integer>m=new HashMap();for(String w:b.toString().toLowerCase().split("(\\b(.|the|and|of|to|i[tns]|or)\\b|\\W)+"))m.put(w,m.get(w)!=null?m.get(w)+1:1);
  List<String>l=new Vector(m.keySet());Collections.sort(l,new Comparator(){public int compare(Object l,Object r){return m.get(r)-m.get(l);}});
  int c=76-l.get(0).length();String s=new String(new char[c]).replace('\0','_');System.out.println(" "+s);
  for(String w:l.subList(0,22))System.out.println("|"+s.substring(0,m.get(w)*c/m.get(l.get(0)))+"| "+w);
 }
}

Output:

 _________________________________________________________________________
|_________________________________________________________________________| she
|_______________________________________________________________| you
|____________________________________________________________| said
|_____________________________________________________| alice
|_______________________________________________| was
|___________________________________________| that
|____________________________________| as
|________________________________| her
|_____________________________| with
|_____________________________| at
|__________________________| on
|__________________________| all
|_______________________| this
|_______________________| for
|_______________________| had
|_______________________| but
|______________________| be
|_____________________| not
|____________________| they
|____________________| so
|___________________| very
|__________________| what

It pretty sucks that Java doesn't have String#join() and closures (yet).

Edit by Rotsor:

I have made several changes to your solution:

• Replaced List with a String[]
• Reused the 'args' argument instead of declaring my own String array. Also used it as an argument to .ToArray()
• Replaced StringBuffer with a String (yes, yes, terrible performance)
• Replaced Java sorting with a selection-sort with early halting (only first 22 elements have to be found)
• Aggregated some int declaration into a single statement
• Implemented the non-cheating algorithm finding the most limiting line of output. Implemented it without FP.
• Fixed the problem of the program crashing when there were less than 22 distinct words in the text
• Implemented a new algorithm of reading input, which is fast and only 9 characters longer than the slow one.

The condensed code is 688 711 684 characters long:

import java.util.*;class F{public static void main(String[]l)throws Exception{Map<String,Integer>m=new HashMap();String w="";int i=0,k=0,j=8,x,y,g=22;for(;(j=System.in.read())>0;w+=(char)j);for(String W:w.toLowerCase().split("(\\b(.|the|and|of|to|i[tns]|or)\\b|\\W)+"))m.put(W,m.get(W)!=null?m.get(W)+1:1);l=m.keySet().toArray(l);x=l.length;if(x<g)g=x;for(;i<g;++i)for(j=i;++j<x;)if(m.get(l[i])<m.get(l[j])){w=l[i];l[i]=l[j];l[j]=w;}for(;k<g;k++){x=76-l[k].length();y=m.get(l[k]);if(k<1||y*i>x*j){i=x;j=y;}}String s=new String(new char[m.get(l[0])*i/j]).replace('\0','_');System.out.println(" "+s);for(k=0;k<g;k++){w=l[k];System.out.println("|"+s.substring(0,m.get(w)*i/j)+"| "+w);}}}

The fast version (720 693 characters)

import java.util.*;class F{public static void main(String[]l)throws Exception{Map<String,Integer>m=new HashMap();String w="";int i=0,k=0,j=8,x,y,g=22;for(;j>0;){j=System.in.read();if(j>90)j-=32;if(j>64&j<91)w+=(char)j;else{if(!w.matches("^(|.|THE|AND|OF|TO|I[TNS]|OR)$"))m.put(w,m.get(w)!=null?m.get(w)+1:1);w="";}}l=m.keySet().toArray(l);x=l.length;if(x<g)g=x;for(;i<g;++i)for(j=i;++j<x;)if(m.get(l[i])<m.get(l[j])){w=l[i];l[i]=l[j];l[j]=w;}for(;k<g;k++){x=76-l[k].length();y=m.get(l[k]);if(k<1||y*i>x*j){i=x;j=y;}}String s=new String(new char[m.get(l[0])*i/j]).replace('\0','_');System.out.println(" "+s);for(k=0;k<g;k++){w=l[k];System.out.println("|"+s.substring(0,m.get(w)*i/j)+"| "+w);}}}

More readable version:

import java.util.*;class F{public static void main(String[]l)throws Exception{
    Map<String,Integer>m=new HashMap();String w="";
    int i=0,k=0,j=8,x,y,g=22;
    for(;j>0;){j=System.in.read();if(j>90)j-=32;if(j>64&j<91)w+=(char)j;else{
        if(!w.matches("^(|.|THE|AND|OF|TO|I[TNS]|OR)$"))m.put(w,m.get(w)!=null?m.get(w)+1:1);w="";
    }}
    l=m.keySet().toArray(l);x=l.length;if(x<g)g=x;
    for(;i<g;++i)for(j=i;++j<x;)if(m.get(l[i])<m.get(l[j])){w=l[i];l[i]=l[j];l[j]=w;}
    for(;k<g;k++){x=76-l[k].length();y=m.get(l[k]);if(k<1||y*i>x*j){i=x;j=y;}}
    String s=new String(new char[m.get(l[0])*i/j]).replace('\0','_');
    System.out.println(" "+s);
    for(k=0;k<g;k++){w=l[k];System.out.println("|"+s.substring(0,m.get(w)*i/j)+"| "+w);}}
}

The version without behaviour improvements is 615 characters:

import java.util.*;class F{public static void main(String[]l)throws Exception{Map<String,Integer>m=new HashMap();String w="";int i=0,k=0,j=8,g=22;for(;j>0;){j=System.in.read();if(j>90)j-=32;if(j>64&j<91)w+=(char)j;else{if(!w.matches("^(|.|THE|AND|OF|TO|I[TNS]|OR)$"))m.put(w,m.get(w)!=null?m.get(w)+1:1);w="";}}l=m.keySet().toArray(l);for(;i<g;++i)for(j=i;++j<l.length;)if(m.get(l[i])<m.get(l[j])){w=l[i];l[i]=l[j];l[j]=w;}i=76-l[0].length();String s=new String(new char[i]).replace('\0','_');System.out.println(" "+s);for(k=0;k<g;k++){w=l[k];System.out.println("|"+s.substring(0,m.get(w)*i/m.get(l[0]))+"| "+w);}}}

Answer 3

Perl: 203 202 201 198 195 208 203 / 231 chars

$/=\0;/^(the|and|of|to|.|i[tns]|or)$/i||$x{lc$_}++for<>=~/[a-z]+/gi;map{$z=$x{$_};$y||{$y=(76-y///c)/$z}&&warn" "."_"x($z*$y)."\n";printf"|%.78s\n","_"x($z*$y)."| $_"}(sort{$x{$b}<=>$x{$a}}keys%x)[0..21]

Alternate, full implementation including indicated behaviour (global bar-squishing) for the pathological case in which the secondary word is both popular and long enough to combine to over 80 chars (this implementation is 231 chars):

$/=\0;/^(the|and|of|to|.|i[tns]|or)$/i||$x{lc$_}++for<>=~/[a-z]+/gi;@e=(sort{$x{$b}<=>$x{$a}}keys%x)[0..21];for(@e){$p=(76-y///c)/$x{$_};($y&&$p>$y)||($y=$p)}warn" "."_"x($x{$e[0]}*$y)."\n";for(@e){warn"|"."_"x($x{$_}*$y)."| $_\n"}

The specification didn't state anywhere that this had to go to STDOUT, so I used perl's warn() instead of print - four characters saved there. Used map instead of foreach, but I feel like there could still be some more savings in the split(join()). Still, got it down to 203 - might sleep on it. At least Perl's now under the "shell, grep, tr, grep, sort, uniq, sort, head, perl" char count for now ;)

PS: Reddit says "Hi" ;)

Update: Removed join() in favour of assignment and implicit scalar conversion join. Down to 202. Also please note I have taken advantage of the optional "ignore 1-letter words" rule to shave 2 characters off, so bear in mind the frequency count will reflect this.

Update 2: Swapped out assignment and implicit join for killing $/ to get the file in one gulp using <> in the first place. Same size, but nastier. Swapped out if(!$y){} for $y||{}&&, saved 1 more char => 201.

Update 3: Took control of lowercasing early (lc<>) by moving lc out of the map block - Swapped out both regexes to no longer use /i option, as no longer needed. Swapped explicit conditional x?y:z construct for traditional perlgolf || implicit conditional construct - /^...$/i?1:$x{$}++ for /^...$/||$x{$}++ Saved three characters! => 198, broke the 200 barrier. Might sleep soon... perhaps.

Update 4: Sleep deprivation has made me insane. Well. More insane. Figuring that this only has to parse normal happy text files, I made it give up if it hits a null. Saved two characters. Replaced "length" with the 1-char shorter (and much more golfish) y///c - you hear me, GolfScript?? I'm coming for you!!! sob

Update 5: Sleep dep made me forget about the 22row limit and subsequent-line limiting. Back up to 208 with those handled. Not too bad, 13 characters to handle it isn't the end of the world. Played around with perl's regex inline eval, but having trouble getting it to both work and save chars... lol. Updated the example to match current output.

Update 6: Removed unneeded braces protecting (...)for, since the syntactic candy ++ allows shoving it up against the for happily. Thanks to input from Chas. Owens (reminding my tired brain), got the character class i[tns] solution in there. Back down to 203.

Update 7: Added second piece of work, full implementation of specs (including the full bar-squishing behaviour for secondary long-words, instead of truncation which most people are doing, based on the original spec without the pathological example case)

Examples:

 _________________________________________________________________________
|_________________________________________________________________________| she
|_______________________________________________________________| you
|____________________________________________________________| said
|_____________________________________________________| alice
|_______________________________________________| was
|___________________________________________| that
|____________________________________| as
|________________________________| her
|_____________________________| with
|_____________________________| at
|__________________________| on
|__________________________| all
|_______________________| this
|_______________________| for
|_______________________| had
|_______________________| but
|______________________| be
|_____________________| not
|____________________| they
|____________________| so
|___________________| very
|__________________| what

Alternative implementation in pathological case example:

 _______________________________________________________________
|_______________________________________________________________| she
|_______________________________________________________| superlongstringstring
|____________________________________________________| said
|______________________________________________| alice
|________________________________________| was
|_____________________________________| that
|_______________________________| as
|____________________________| her
|_________________________| with
|_________________________| at
|_______________________| on
|______________________| all
|____________________| this
|____________________| for
|____________________| had
|____________________| but
|___________________| be
|__________________| not
|_________________| they
|_________________| so
|________________| very
|________________| what

Answer 4

Gawk -- 336 (originally 507) characters

(after fixing the output formatting; fixing the contractions thing; tweaking; tweaking again; removing a wholly unnecessary sorting step; tweaking yet again; and again (oops this one broke the formatting); tweak some more; taking up Matt's challenge I desperately tweak so more; found another place to save a few, but gave two back to fix the bar length bug)

Heh heh! I am momentarily ahead of [Matt's JavaScript][1] solution^{counter challenge! ;)} and [AKX's python][2].

The problem seems to call out for a language that implements native associative arrays, so of course I've chosen one with a horribly deficient set of operators on them. In particular, you cannot control the order in which awk offers up the elements of a hash map, so I repeatedly scan the whole map to find the currently most numerous item, print it and delete it from the array.

It is all terribly inefficient, with all the golfifcations I've made it has gotten to be pretty awful, as well.

Minified:

{gsub("[^a-zA-Z]"," ");for(;NF;NF--)a[tolower($NF)]++}
END{split("the and of to a i it in or is",b," ");
for(w in b)delete a[b[w]];d=1;for(w in a){e=a[w]/(78-length(w));if(e>d)d=e}
for(i=22;i;--i){e=0;for(w in a)if(a[w]>e)e=a[x=w];l=a[x]/d-2;
t=sprintf(sprintf("%%%dc",l)," ");gsub(" ","_",t);if(i==22)print" "t;
print"|"t"| "x;delete a[x]}}

line breaks for clarity only: they are not necessary and should not be counted.

---

Output:

$ gawk -f wordfreq.awk.min < 11.txt 
 _________________________________________________________________________
|_________________________________________________________________________| she
|_______________________________________________________________| you
|____________________________________________________________| said
|____________________________________________________| alice
|______________________________________________| was
|__________________________________________| that
|___________________________________| as
|_______________________________| her
|____________________________| with
|____________________________| at
|___________________________| s
|___________________________| t
|_________________________| on
|_________________________| all
|______________________| this
|______________________| for
|______________________| had
|_____________________| but
|____________________| be
|____________________| not
|___________________| they
|__________________| so
$ sed 's/you/superlongstring/gI' 11.txt | gawk -f wordfreq.awk.min
 ______________________________________________________________________
|______________________________________________________________________| she
|_____________________________________________________________| superlongstring
|__________________________________________________________| said
|__________________________________________________| alice
|____________________________________________| was
|_________________________________________| that
|_________________________________| as
|______________________________| her
|___________________________| with
|___________________________| at
|__________________________| s
|__________________________| t
|________________________| on
|________________________| all
|_____________________| this
|_____________________| for
|_____________________| had
|____________________| but
|___________________| be
|___________________| not
|__________________| they
|_________________| so

Readable; 633 characters (originally 949):

{
    gsub("[^a-zA-Z]"," ");
    for(;NF;NF--)
    a[tolower($NF)]++
}
END{
    # remove "short" words
    split("the and of to a i it in or is",b," ");
    for (w in b) 
    delete a[b[w]];
    # Find the bar ratio
    d=1;
    for (w in a) {
    e=a[w]/(78-length(w));
    if (e>d)
        d=e
    }
    # Print the entries highest count first
    for (i=22; i; --i){               
    # find the highest count
    e=0;
    for (w in a) 
        if (a[w]>e)
        e=a[x=w];
        # Print the bar
    l=a[x]/d-2;
    # make a string of "_" the right length
    t=sprintf(sprintf("%%%dc",l)," ");
    gsub(" ","_",t);
    if (i==22) print" "t;
    print"|"t"| "x;
    delete a[x]
    }
}

Answer 5

Haskell - 366 351 344 337 333 characters

(One line break in main added for readability, and no line break needed at end of last line.)

import Data.List
import Data.Char
l=length
t=filter
m=map
f c|isAlpha c=toLower c|0<1=' '
h w=(-l w,head w)
x!(q,w)='|':replicate(minimum$m(q?)x)'_'++"| "++w
q?(g,w)=q*(77-l w)`div`g
b x=m(x!)x
a(l:r)=(' ':t(=='_')l):l:r
main=interact$unlines.a.b.take 22.sort.m h.group.sort
  .t(`notElem`words"the and of to a i it in or is").words.m f

How it works is best seen by reading the argument to interact backwards:

• map f lowercases alphabetics, replaces everything else with spaces.
• words produces a list of words, dropping the separating whitespace.
• filter (notElemwords "the and of to a i it in or is") discards all entries with forbidden words.
• group . sort sorts the words, and groups identical ones into lists.
• map h maps each list of identical words to a tuple of the form (-frequency, word).
• take 22 . sort sorts the tuples by descending frequency (the first tuple entry), and keeps only the first 22 tuples.
• b maps tuples to bars (see below).
• a prepends the first line of underscores, to complete the topmost bar.
• unlines joins all these lines together with newlines.

The tricky bit is getting the bar length right. I assumed that only underscores counted towards the length of the bar, so || would be a bar of zero length. The function b maps c x over x, where x is the list of histograms. The entire list is passed to c, so that each invocation of c can compute the scale factor for itself by calling u. In this way, I avoid using floating-point math or rationals, whose conversion functions and imports would eat many characters.

Note the trick of using -frequency. This removes the need to reverse the sort since sorting (ascending) -frequency will places the words with the largest frequency first. Later, in the function u, two -frequency values are multiplied, which will cancel the negation out.

Answer 6

PHP CLI version (450 chars)

This solution takes into account the last requirement which most purists have conviniently chosen to ignore. That costed 170 characters!

Usage: php.exe <this.php> <file.txt>

Minified:

<?php $a=array_count_values(array_filter(preg_split('/[^a-z]/',strtolower(file_get_contents($argv[1])),-1,1),function($x){return !preg_match("/^(.|the|and|of|to|it|in|or|is)$/",$x);}));arsort($a);$a=array_slice($a,0,22);function R($a,$F,$B){$r=array();foreach($a as$x=>$f){$l=strlen($x);$r[$x]=$b=$f*$B/$F;if($l+$b>76)return R($a,$f,76-$l);}return$r;}$c=R($a,max($a),76-strlen(key($a)));foreach($a as$x=>$f)echo '|',str_repeat('-',$c[$x]),"| $x\n";?>

Human readable:

<?php

// Read:
$s = strtolower(file_get_contents($argv[1]));

// Split:
$a = preg_split('/[^a-z]/', $s, -1, PREG_SPLIT_NO_EMPTY);

// Remove unwanted words:
$a = array_filter($a, function($x){
       return !preg_match("/^(.|the|and|of|to|it|in|or|is)$/",$x);
     });

// Count:
$a = array_count_values($a);

// Sort:
arsort($a);

// Pick top 22:
$a=array_slice($a,0,22);


// Recursive function to adjust bar widths
// according to the last requirement:
function R($a,$F,$B){
    $r = array();
    foreach($a as $x=>$f){
        $l = strlen($x);
        $r[$x] = $b = $f * $B / $F;
        if ( $l + $b > 76 )
            return R($a,$f,76-$l);
    }
    return $r;
}

// Apply the function:
$c = R($a,max($a),76-strlen(key($a)));


// Output:
foreach ($a as $x => $f)
    echo '|',str_repeat('-',$c[$x]),"| $x\n";

?>

Output:

|-------------------------------------------------------------------------| she
|---------------------------------------------------------------| you
|------------------------------------------------------------| said
|-----------------------------------------------------| alice
|-----------------------------------------------| was
|-------------------------------------------| that
|------------------------------------| as
|--------------------------------| her
|-----------------------------| at
|-----------------------------| with
|--------------------------| on
|--------------------------| all
|-----------------------| this
|-----------------------| for
|-----------------------| had
|-----------------------| but
|----------------------| be
|---------------------| not
|--------------------| they
|--------------------| so
|-------------------| very
|------------------| what

When there is a long word, the bars are adjusted properly:

|--------------------------------------------------------| she
|---------------------------------------------------| thisisareallylongwordhere
|-------------------------------------------------| you
|-----------------------------------------------| said
|-----------------------------------------| alice
|------------------------------------| was
|---------------------------------| that
|---------------------------| as
|-------------------------| her
|-----------------------| with
|-----------------------| at
|--------------------| on
|--------------------| all
|------------------| this
|------------------| for
|------------------| had
|-----------------| but
|-----------------| be
|----------------| not
|---------------| they
|---------------| so
|--------------| very