LZW - again

This is do with LZW.

I am trying to read in codewords as dictionary entries from a file, I read in 7 bits to start with, because 0 - 127
is the number of original entries in my dictionary ( i.e all the ascii values from 0 - 127 ), but when reading bits
from a file, i read each bit at a time but each bit ( up to 8 ) is stored in one byte, however if a number such as
one greater than 255 it will be 9 bits long, so that i need to be able to recognise this.  Note that when I compress
I have a current length variable that changes if the binary string that i am about to compress is greater than
its current length, if so i change current length to the size of that string, so that any binary strings (numbers)
processed after the change that are shorter in length than the current length are padded with 0's at the begining i.e

take the following
1001001                    //curlength variable is 7
1000101
10010101                //curlength variable is now changed to 8 as string length has changed
1111                 //but now we have a binary string of 4 length
 will become

1001001              //curlength variable is 7
1000101
10010101               //curlength variable is now changed to 8 as string length has changed
00001111               //curlength is no longer 7 so pad four 0's to make it the length of curlength variable


So when decompressing I will have some codewords padded with 0.

if i reach the number 127 at any point then i change the value of the curlength to 8, if
i reach the value of 255 at any time, then i change the value of the curlength 255 and so on, so that
even if get a code word that would be less than that then i know that it will be padded
with 0s at the begining.  The problem is is how do i know when to change as I am reading in 8 bits at a time
so that if a certain number is 9 bits what do I do?

I can post code if that will help, any help greatly appreciated.