Spell Check Solution

Your task is to write a Multi-Threaded (ie. concurrent) Client Server C/C++ application for performing spell checking of text files using the dictionary in /usr/dict/words.

After the file is processed by the server, the client should display all miss-spelt words and any near words as suggested replacements. (An example output is shown below.)
Line No. Mis-spelt word Suggestion
7 enogh enough
12 nothng nothing
230 happenned happened
234 glob globe
563 poeple people
1034 lathgter laughter
1455 kowing knowing
1788 wuold would
1951 quate quite
2145 javascript no suggestion
2344 winkle wrinkle
Total processing time = 3.723 sec
File processing should be done by the server on a line by line basis. Each time a line of text is read it should be preprocessed by converting all uppercase letters to lowercase and removing leading or tailing punctuation characters. A word is to be defined as a sequence alphabetic characters as well as the characters ' ' ' (apostrophe) and '-'(hyphen). Thus, words like "co-operate" and "it's" will be processed as whole words. (Note: If you are not sure how to extract words from a line of text you can use the example code in sscanf_eg.cpp.) The total processing time is the time from the commencement of the program (or client) until its termination. To search for words in the dictionary just use a linear search.

To determine if the two words are the same (or different) use the int WordCmp(char*,char*) function below. It returns 0 if the words are the same or a number 0 if the words are different. (Note: the greater the return value the greater the difference between the words.) To obtain a suggestion word for a miss-spelt word use the word with the minimum difference. (Note: if there is more than one word with the minimum difference for a miss-spelt word then use the one found first as the suggestion word.)
int WordCmp(char*Word1,char*Word2){
return 0;
return 100;
float AveWordLen = (strlen(Word1) + strlen(Word2)) / 2.0;
return int(NumUniqueChars(Word1,Word2) / AveWordLen * 100);
where: the `NumUniqueChars()` is the number of chars in Word1 that are not in Word2 plus the number of chars in Word2 that are not in Word1. You may improve this compare function if you wish.

**You should complete this program by performing the following steps.**

Step 1: Implement a single-process program without sockets or threads.

Step 2: Convert the program in step 1 into a client-server application using TCP sockets.

Step 3: Convert the server program in step 2 into a multithreaded program capable of processing lines of
input file data in parallel from an input queue and record all miss-spelt words, their suggestion
word and line number in a global linked list. You may use buffer.cc to help with this step. The
mutexes and condition vars should provide mutual exclusion to the shared global data to prevent
the threads from updating it simultaneously. When processing is complete the contents of the
linked list are sent to the client for displaying on the user's screen.


open input file
while !eof
read line from file
preprocess line
for each word in line
open dictionary
for each word in dictionary
compare words
if same found=true;
if word difference is minimum
keep dictionary word as a suggestion word
close dictionary
if !found
print line no. word and suggestion on screen
end for
close input file
print processing time

1. get filename and number of slaves etc. from command line
2. gethostbyname (to obtain IP_Address of the Server)
3. contact Server (using connect + IP_Address from above)
4. send filename and number of threads to server & wait for miss-spelt words to return.
5. take miss-spelt words from Server and print them
6. close socket.


1. setup socket
2. accept client connection
4. read filename and number of threaded slaves.
5. open file
6. create threaded slaves
7. keep filling input buffer with preprocessed lines of text. (see Producer below)
8. wait for threads to complete
9. send miss-spelt words to client
10. close socket & file, goto 2

read line from file
if no data left set data_finished & break;
if queue full pthread_cond_wait
put item in input queue
. . .
if data_finished break;
if input queue empty pthread_cond_wait
take item from queue
process item
. . .

(i) For the server to work correctly you will have to implement a number of mutexes and condition
variables appropriately to make the global data accesses mutually exclusive. Remember with threads
preemption can occur at any point in the program, even halfway through calculations.

(ii) Using the following attribute ensures that threads will be bound to kernel LWPs in a 1:1 manner - ie. the OS uses time slicing (this ensures that each thread will get CPU time allocated, & overcomes potential Solaris thread scheduling problems):
pthread_attr_setscope(&attr, PTHREAD_SCOPE_SYSTEM);

(iii) Compile using the -mt & -lpthread options \[ref. cc man page entries\]
(By default, -mt appends -lthread ie. Solaris rather than POSIX!)
Powered by