Implement a program that spell-checks a file, per the below.

$ ./speller texts/austinpowers.txt


TIME IN load:
TIME IN check:
TIME IN size:
TIME IN unload:



$ wget https://github.com/cs50/problems/archive/speller.zip
$ unzip speller.zip
$ rm speller.zip
$ mv problems-speller speller
$ cd speller
$ ls
dictionaries/  dictionary.c  dictionary.h  keys/  Makefile  questions.txt  speller.c  texts/


Theoretically, on input of size n, an algorithm with a running time of n is asymptotically equivalent, in terms of O, to an algorithm with a running time of 2n. In the real world, though, the fact of the matter is that the latter feels twice as slow as the former.

The challenge ahead of you is to implement the fastest spell checker you can! By "fastest," though, we’re talking actual, real-world, noticeable time—none of that asymptotic stuff this time.

In speller.c, we’ve put together a program that’s designed to spell-check a file after loading a dictionary of words from disk into memory. Unfortunately, we didn’t quite get around to implementing the loading part. Or the checking part. Both (and a bit more) we leave to you! But first, a tour.


Open up dictionary.h. Declared in that file are four functions; take note of what each should do. Now open up dictionary.c. Notice that we’ve implemented those four functions, but only barely, just enough for this code to compile. Your job, ultimately, is to re-implement those functions as cleverly as possible so that this spell checker works as advertised. And fast!


Recall that make automates compilation of your code so that you don’t have to execute clang manually along with a whole bunch of switches. However, as your programs grow in size, make won’t be able to infer from context anymore how to compile your code; you’ll need to start telling make how to compile your program, particularly when they involve multiple source (i.e., .c) files, as in the case of this problem. And so we’ll utilize a Makefile, a configuration file that tells make exactly what to do. Open up Makefile, and let’s take a tour of its lines.

The line below defines a variable called CC that specifies that make should use clang for compiling.

CC = clang

The line below defines a variable called CFLAGS that specifies, in turn, that clang should use some flags, most of which should look familiar.

CFLAGS = -ggdb3 -O0 -Qunused-arguments -std=c11 -Wall -Werror

The line below defines a variable called EXE, the value of which will be our program’s name.

EXE = speller

The line below defines a variable called HDRS, the value of which is a space-separated list of header files used by speller.

HDRS = dictionary.h

The line below defines a variable called LIBS, the value of which is should be a space-separated list of libraries, each of which should be prefixed with -l. (Recall our use of -lcs50 earlier this term.) Odds are you won’t need to enumerate any libraries for this problem, but we’ve included the variable just in case.


The line below defines a variable called SRCS, the value of which is a space-separated list of C files that will collectively implement speller.

SRCS = speller.c dictionary.c

The line below defines a variable called OBJS, the value of which is identical to that of SRCS, except that each file’s extension is not .c but .o.

OBJS = $(SRCS:.c=.o)

The lines below define a "target" using these variables that tells make how to compile speller.

$(EXE): $(OBJS) Makefile
    $(CC) $(CFLAGS) -o $@ $(OBJS) $(LIBS)

The line below specifies that our .o files all "depend on" dictionary.h and Makefile so that changes to either induce recompilation of the former when you run make.

$(OBJS): $(HDRS) Makefile

Finally, the lines below define another target for cleaning up this problem’s directory.

    rm -f core $(EXE) *.o

Know that you’re welcome to modify this Makefile as you see fit. In fact, you should if you create any .c or .h files of your own. But be sure not to change any tabs (i.e., \t) to spaces, since make expects the former to be present below each target.

The net effect of all these lines is that you can compile speller with a single command, even though it comprises quite a few files:

make speller

Even better, you can also just execute:


And if you ever want to delete speller plus any core or .o files, you can do so with a single command:

make clean

In general, though, anytime you want to compile your code for this problem, it should suffice to run:



Okay, next open up speller.c and spend some time looking over the code and comments therein. You won’t need to change anything in this file, but you should understand it nonetheless. Notice how, by way of getrusage, we’ll be "benchmarking" (i.e., timing the execution of) your implementations of check, load, size, and unload. Also notice how we go about passing check, word by word, the contents of some file to be spell-checked. Ultimately, we report each misspelling in that file along with a bunch of statistics.

Notice, incidentally, that we have defined the usage of speller to be

Usage: speller [dictionary] text

where dictionary is assumed to be a file containing a list of lowercase words, one per line, and text is a file to be spell-checked. As the brackets suggest, provision of dictionary is optional; if this argument is omitted, speller will use dictionaries/large by default. In other words, running

./speller text

will be equivalent to running

./speller dictionaries/large text

where text is the file you wish to spell-check. Suffice it to say, the former is easier to type! (Of course, speller will not be able to load any dictionaries until you implement load in dictionary.c! Until then, you’ll see Could not load.)

Within the default dictionary, mind you, are 143,091 words, all of which must be loaded into memory! In fact, take a peek at that file to get a sense of its structure and size. Notice that every word in that file appears in lowercase (even, for simplicity, proper nouns and acronyms). From top to bottom, the file is sorted lexicographically, with only one word per line (each of which ends with \n). No word is longer than 45 characters, and no word appears more than once. During development, you may find it helpful to provide speller with a dictionary of your own that contains far fewer words, lest you struggle to debug an otherwise enormous structure in memory. In dictionaries/small is one such dictionary. To use it, execute

./speller dictionaries/small text

where text is the file you wish to spell-check. Don’t move on until you’re sure you understand how speller itself works!

Odds are, you didn’t spend enough time looking over speller.c. Go back one square and walk yourself through it again!


So that you can test your implementation of speller, we’ve also provided you with a whole bunch of texts, among them the script from Austin Powers: International Man of Mystery, a sound bite from Ralph Wiggum, three million bytes from Tolstoy, some excerpts from Machiavelli and Shakespeare, the entirety of the King James V Bible, and more. So that you know what to expect, open and skim each of those files, all of which are in a directory called texts within your pset5 directory.

Now, as you should know from having read over speller.c carefully, the output of speller, if executed with, say,

./speller texts/austinpowers.txt

will eventually resemble the below. For now, try executing the staff’s solution (using the default dictionary) with the below.

~cs50/pset5/speller texts/austinpowers.txt

Below’s some of the output you’ll see. For amusement’s sake, we’ve excerpted some of our favorite "misspellings." And lest we spoil the fun, we’ve omitted our own statistics for now.



TIME IN load:
TIME IN check:
TIME IN size:
TIME IN unload:

TIME IN load represents the number of seconds that speller spends executing your implementation of load. TIME IN check represents the number of seconds that speller spends, in total, executing your implementation of check. TIME IN size represents the number of seconds that speller spends executing your implementation of size. TIME IN unload represents the number of seconds that speller spends executing your implementation of unload. TIME IN TOTAL is the sum of those four measurements.

Note that these times may vary somewhat across executions of speller, depending on what else CS50 IDE is doing, even if you don’t change your code.

Incidentally, to be clear, by "misspelled" we simply mean that some word is not in the dictionary provided.


Open up questions.txt and answer each of the following questions in one or more sentences.

  1. What is pneumonoultramicroscopicsilicovolcanoconiosis?

  2. According to its man page, what does getrusage do?

  3. Per that same man page, how many members are in a variable of type struct rusage?

  4. Why do you think we pass before and after by reference (instead of by value) to calculate, even though we’re not changing their contents?

  5. Explain as precisely as possible, in a paragraph or more, how main goes about reading words from a file. In other words, convince us that you indeed understand how that function’s for loop works.

  6. Why do you think we used fgetc to read each word’s characters one at a time rather than use fscanf with a format string like "%s" to read whole words at a time? Put another way, what problems might arise by relying on fscanf alone?

  7. Why do you think we declared the parameters for check and load as const (which means "constant")?


Alright, the challenge now before you is to implement load, check, size, and unload as efficiently as possible, in such a way that TIME IN load, TIME IN check, TIME IN size, and TIME IN unload are all minimized. To be sure, it’s not obvious what it even means to be minimized, inasmuch as these benchmarks will certainly vary as you feed speller different values for dictionary and for text. But therein lies the challenge, if not the fun, of this problem. This problem is your chance to design. Although we invite you to minimize space, your ultimate enemy is time. But before you dive in, some specifications from us.

  • You may not alter speller.c.

  • You may alter dictionary.c (and, in fact, must in order to complete the implementations of load, check, size, and unload), but you may not alter the declarations of load, check, size, or unload.

  • You may alter dictionary.h, but you may not alter the declarations of load, check, size, or unload.

  • You may alter Makefile.

  • You may add functions to dictionary.c or to files of your own creation so long as all of your code compiles via make.

  • Your implementation of check must be case-insensitive. In other words, if foo is in dictionary, then check should return true given any capitalization thereof; none of foo, foO, fOo, fOO, fOO, Foo, FoO, FOo, and FOO should be considered misspelled.

  • Capitalization aside, your implementation of check should only return true for words actually in dictionary. Beware hard-coding common words (e.g., the), lest we pass your implementation a dictionary without those same words. Moreover, the only possessives allowed are those actually in dictionary. In other words, even if foo is in dictionary, check should return false given foo’s if foo’s is not also in dictionary.

  • You may assume that check will only be passed strings with alphabetical characters and/or apostrophes.

  • You may assume that any dictionary passed to your program will be structured exactly like ours, lexicographically sorted from top to bottom with one word per line, each of which ends with \n. You may also assume that dictionary will contain at least one word, that no word will be longer than LENGTH (a constant defined in dictionary.h) characters, that no word will appear more than once, and that each word will contain only lowercase alphabetical characters and possibly apostrophes.

  • Your spell checker may only take text and, optionally, dictionary as input. Although you might be inclined (particularly if among those more comfortable) to "pre-process" our default dictionary in order to derive an "ideal hash function" for it, you may not save the output of any such pre-processing to disk in order to load it back into memory on subsequent runs of your spell checker in order to gain an advantage.

  • Your spell checker may not leak any memory.

  • You may search for (good) hash functions online, so long as you cite the origin of any hash function you integrate into your own code.

Alright, ready to go?

  1. Implement load.

  2. Implement check.

  3. Implement size.

  4. Implement unload.



Be sure to free in unload any memory that you allocated in load! Recall that valgrind is your newest best friend. Know that valgrind watches for leaks while your program is actually running, so be sure to provide command-line arguments if you want valgrind to analyze speller while you use a particular dictionary and/or text, as in the below. Best to use a small text, though, else valgrind could take quite a while to run.

valgrind --leak-check=full ./speller texts/ralph.txt

If you run valgrind without specifying a text for speller, your implementations of load and unload won’t actually get called (and thus analyzed).

If unsure how to interpret the output of valgrind, do just ask help50 for help:

help50 valgrind --leak-check=full ./speller texts/ralph.txt


How to check whether your program is outting the right misspelled words? Well, you’re welcome to consult the "answer keys" that are inside of the keys directory that’s inside of your speller directory. For instance, inside of keys/austinpowers.txt are all of the words that your program should think are misspelled.

You could therefore run your program on some text in one window, as with the below.

./speller texts/austinpowers.txt

And you could then run the staff’s solution on the same text in another window, as with the below.

~cs50/pset5/speller texts/austinpowers.txt

And you could then compare the windows visually side by side. That could get tedious quickly, though. So you might instead want to "redirect" your program’s output to a file (just like you may have done with generate in Problem Set 3), as with the below.

./speller texts/austinpowers.txt > student.txt
~cs50/pset5/speller texts/austinpowers.txt > staff.txt

You can then compare both files side by side in the same window with a program like diff, as with the below.

diff -y student.txt staff.txt

Alternatively, to save time, you could just compare your program’s output (assuming you redirected it to, e.g., student.txt) against one of the answer keys without running the staff’s solution, as with the below.

diff -y student.txt keys/austinpowers.txt

If your program’s output matches the staff’s, diff will output two columns that should be identical except for, perhaps, the running times at the bottom. If the columns differ, though, you’ll see a > or | where they differ. For instance, if you see

MISSPELLED WORDS                                                MISSPELLED WORDS

FOTTAGE                                                         FOTTAGE
INT                                                             INT
                                                              > EVIL'S
s                                                               s
                                                              > EVIL'S
Farbissina                                                      Farbissina

that means your program (whose output is on the left) does not think that EVIL’s is misspelled, even though the staff’s output (on the right) does, as is implied by the absence of EVIL’s in the lefthand column and the presence of EVIL’s in the righthand column.


To test your code less manually (though still not exhaustively), you may also execute the below.

check50 cs50/2017/x/speller

Note that check50 will also check for memory leaks, so be sure you’ve run valgrind as well.

Staff’s Solution

How to assess just how fast (and correct) your code is? Well, as always, feel free to play with the staff’s solution, as with the below, and compare its numbers against yours.

~cs50/pset5/speller texts/austinpowers.txt


None so far! Reload this page periodically to check if any arise!


  • 2016-10-08

    • Corrected walkthrough for load, wherein sample code had

      node1->word = "Hello";
      node2->word = "World";

      instead of

      strcpy(node1->word, "Hello");
      strcpy(node2->word, "World");
  • 2016-09-30

    • Initial release.