I have a large text file to be opened (eg- 5GB size). But with a limited RAM (take 1 GB), How can I open and read the file with out any memory error? I am running on a linux terminal with with the basic packages installed.
This was an interview question, hence please do not look into the practicality.
I do not know whether to look at it in System level or programmatic level... It would be great if someone can throw some light into this issue.
Thanks.
Read it character by character... or X bytes by X bytes... it really depends what you want to do with it... As long as you don't need the whole file at once, that works.
(Ellipses are awesome)
What do they want you to do with the file? Are you looking for something? Extracting something? Sorting? This will affect your approach.
It may be sufficient to read the file line by line or character by character if you're looking for something. If you need to jump around the file or analyze sections of it, then most likely want to memory map it. Look up mmap(). Here's an short article on the subject:memory mapped i/o
[just comment]
If you are going to use system calls (open() and read()), then reading character by character will generate a lot of system calls that severely slow down your application. Even with the existence of the buffer cache (or disk file), system calls are expensive.
It is better to read block by block where block size "SHOULD" be more than 1MB. In case of 1MB block size, you will issue 5*1024 system calls.
Related
On Linux, is it possible to get the buffer size required for getdents64 to get all the entries in one go (assuming no modifications to the directory after the size is obtained)?
I tried the value from fstat(dirfd,&stb); stb.st_size but it appears unnecessarily too large? What is the value that stat::st_size holds for directories?
As far as I know, no, there is no way to do this, especially not in a manner that works for arbitrary filesystems. The representation of directories is defined by the underlying filesystem, and the same seems to be true of st_size. Certainly there is no way to get the right value for FUSE or 9p or other remote/virtual filesystems.
Why do you want to do this? I don't think it's useful. Once you get beyond a few kB per call, the overhead of the syscalls will be dominated by the actual work done. If you really care you can wrap the getdents64 syscall with a function that keeps resizing the buffer and calling it until EOF is reached. Or you could just use the portable opendir/readdir interfaces which perform appropriate buffering for you.
Box: Linux, gcc.
Problem :
Finding out the file signature of an home folder, which contains thousand of items, by scanning this folder recursively.
Done so far:
Using mmap() system call to load the first 1k byte of each file and check the file magic number.
The drawback with that method is that for each file encountered i've got to make two system calls (e.g mmap() and munmap()).
Best solution if possible:
I would like to allocate a single chunk of memory, load each file (in a row) in this unique buffer and when processing is completed deallocate it, meaning that for each folder scanned i would only use two system calls.
I can't figure out which system call to use in order to achieve that, not even if this solution is realistic!
Any advice would be greatly appreciated.
Don't worry about performance until you know it isn't enough. Your time is much more valuable than the gains in program run time (except for extremely rare cases). And when the performance isn't enough, measure before digging in. There are numerous war stories on "performance optimizations" that were a complete waste (if not positivley harmful).
I have written a C/C++-program for Windows 7 - 64bit that works on very large files. In the final step it reads lines from an input-file (10GB+) and writes them to an output file. The access to the input-file is random, the writing is sequential.
EDIT: Main reason for this approach is to reduce RAM usage.
What I basically do in the reading part is this: (Sorry, very shortened and maybe buggy)
void seekAndGetLine(char* line, size_t lineSize, off64_t pos, FILE* filePointer){
fseeko64(filePointer, pos, ios_base::beg);
fgets(line, lineSize, filePointer);
}
Normally this code is fine, not to say fast, but under some very special conditions it gets very slow. The behaviour doesn't seem to be deterministic, since the performance drops occure on different machines at other parts of the file or even don't occure at all. It even goes so far, that the program totally stops reading, while there are no disc-operations.
Another sympthom seems to be the used RAM. My process keeps it's RAM steady, but the RAM used by the System grows sometimes very large. After using some RAM-Tools I found out, that the Windows Mapped File grows into several GBs. This behaviour also seems to depend on the hardware, since it occure on different machines at different parts of the process.
As far as I can tell, this problem doesn't exist on SSDs, so it definitely has something to do with the responsetime of the HDD.
My guess is that the Windows Caching gets somehow "wierd". The program is fast as long as the cache does it's work. But when Caching goes wrong, the behaviour goes either into "stop reading" or "grow cache size" and sometimes even both. Since I'm no expert for the windows caching algorithms, I would be happy to hear an explanation. Also, is there any way to get Windows out of C/C++ to manipulate/stop/enforce the caching.
Since I'm hunting this problem for a while now, I've already tried some tricks, that didn't work out:
filePointer = fopen(fileName, "rbR"); //Just fills the cache till the RAM is full
massive buffering of the read/write, to stop getting the two into each others way
Thanks in advance
Truly random access across a huge file is the worst possible case for any cache algorithm. It may be best to turn off as much caching as possible.
There are multiple levels of caching:
the CRT library (since you're using the f- functions)
the OS and filesystem
probably onboard the drive itself
If you replace your I/O calls via the f- functions in the CRT with the comparable ones in the Windows API (e.g., CreateFile, ReadFile, etc.) you can eliminate the CRT caching, which may be doing more harm than good. You can also warn the OS that you're going to be doing random accesses, which affects its caching strategy. See options like FILE_FLAG_RANDOM_ACCESS and possibly FILE_FLAG_NO_BUFFERING.
You'll need to experiment and measure.
You might also have to reconsider how your algorithm works. Are the seeks truly random? Can you re-sequence them, perhaps in batches, so that they're in order? Can you limit access to a relatively small region of the file at a time? Can you break the huge file into smaller files and then work with one piece at a time? Have you checked the level of fragmentation on the drive and on the particular file?
Depending on the larger picture of what your application does, you could possibly take a different approach - maybe something like this:
decide which lines you need from the input file and store the
line numbers in a list
sort the list of line numbers
read through the input file once, in order, and pull out the lines
you need (better yet, seek to next line and grab it, especially when there's big gaps)
if the list of lines you're grabbing is small enough, you can store
them in memory for reordering before output, otherwise, stick them
in a smaller temporary file and use that file as input for your
current algorithm to reorder the lines for final output
It's definitely a more complex approach, but it would be much kinder to your caching subsystem, and as a result, could potentially perform significantly better.
I have to analyze a 16 GB file. I am reading through the file sequentially using fread() and fseek(). Is it feasible? Will fread() work for such a large file?
You don't mention a language, so I'm going to assume C.
I don't see any problems with fread, but fseek and ftell may have issues.
Those functions use long int as the data type to hold the file position, rather than something intelligent like fpos_t or even size_t. This means that they can fail to work on a file over 2 GB, and can certainly fail on a 16 GB file.
You need to see how big long int is on your platform. If it's 64 bits, you're fine. If it's 32, you are likely to have problems when using ftell to measure distance from the start of the file.
Consider using fgetpos and fsetpos instead.
Thanks for the response. I figured out where I was going wrong. fseek() and ftell() do not work for files larger than 4GB. I used _fseeki64() and _ftelli64() and it is working fine now.
If implemented correctly this shouldn't be a problem. I assume by sequentially you mean you're looking at the file in discrete chunks and advancing your file pointer.
Check out http://www.computing.net/answers/programming/using-fread-with-a-large-file-/10254.html
It sounds like he was doing nearly the same thing as you.
It depends on what you want to do. If you want to read the whole 16GB of data in memory, then chances are that you'll run out of memory or application heap space.
Rather read the data chunk by chunk and do processing on those chunks (and free resources when done).
But, besides all this, decide which approach you want to do (using fread() or istream, etc.) and do some test cases to see which works better for you.
If you're on a POSIX-ish system, you'll need to make sure you've built your program with 64-bit file offset support. POSIX mandates (or at least allows, and most systems enforce this) the implementation to deny IO operations on files whose size don't fit in off_t, even if the only IO being performed is sequential with no seeking.
On Linux, this means you need to use -D_FILE_OFFSET_BITS=64 on the gcc command line.
I'm writing a little parser and I would like to know the advantages and disadvantages of the different ways to load the data to be parsed. The two ways that I thought of are:
Load the file's contents into a string then parse the string (access the character at an array position)
Parse as reading the file stream (fgetc)
The former will allow me to have two functions: one for parse_from_file and parse_from_string, however I believe this mode will take up more memory. The latter will not have that disadvantage of using more memory.
Does anyone have any advice on the matter?
Reading the entire file in or memory mapping it will be faster, but may cause issues if you want your language to be able to #include other files as these would be memory mapped or read into memory as well.
The stdio functions would work well because they usually try to buffer up data for you, but they are also general purpose so they also try to look out for usage patterns which differ from reading a file from start to finish, but that shouldn't be too much overhead.
A good balance is to have a large circular buffer (x * 2 * 4096 is a good size) which you load with file data and then have your tokenizer read from. Whenever a block's worth of data has been passed to your tokenizer (and you know that it is not going to be pushed back) you can refill that block with new data from the file and update some buffer location info.
Another thing to consider is if there is any chance that the tokenizer would ever need to be able to be used to read from a pipe or from a person typing directly in some text. In these cases your reads may return less data than you asked for without it being at the end of the file, and the buffering method I mentioned above gets more complicated. The stdio buffering is good for this as it can easily be switched to/from line or block buffering (or no buffering).
Using gnu fast lex (flex, but not the Adobe Flash thing) or similar can greatly ease the trouble with all of this. You should look into using it to generate the C code for your tokenizer (lexical analysis).
Whatever you do you should try to make it so that your code can easily be changed to use a different form of next character peek and consume functions so that if you change your mind you won't have to start over.
Consider using lex (and perhaps yacc, if the language of your grammar matches its capabilities). Lex will handle all the fiddly details of lexical analysis for you and produce efficient code. You can probably beat its memory footprint by a few bytes, but how much effort do you want to expend into that?
The most efficient on a POSIX system would probably neither of the two (or a variant of the first if you like): just map the file read-only with mmap, and parse it then. Modern systems are quite efficient with that in that they prefetch data when they detect a streaming access etc., multiple instances of your program that parse the same file will get the same physical pages of memory etc. And the interfaces are relatively simple to handle, I think.