I have a bunch of binary data in N-byte chunks, where each chunk corresponds exactly to one row of a PyTables table.
Right now I am parsing each chunk into fields, writing them to the various fields in the table row, and appending them to the table.
But this seems a little silly since PyTables is going to convert my structured data back into a flat binary form for inclusion in an HDF5 file.
If I need to optimize the CPU time necessary to do this (my data comes in large bursts), is there a more efficient way to load the data into PyTables directly?
PyTables does not currently expose a 'raw' dump mechanism like you describe. However, you can fake it by using UInt8Atom and UInt8Col. You would do something like:
import tables as tb
f = tb.open_file('my_file.h5', 'w')
mytable = f.create_table('/', 'mytable', {'mycol': tb.UInt8Col(shape=(N,))})
mytable.append(myrow)
f.close()
This would likely get you the fastest I/O performance. However, you will miss out on the meaning of the various fields that are part of this binary chunk.
Arguably, raw dumping of the chunks/rows is not what you want to do anyway, which is why it is not explicitly supported. Internally HDF5 and PyTables handle many kinds of conversion for you. This includes but is not limited to things like endianness and thet platform specific feature. By managing the data types for you the resultant HDF5 file and data set cross platform. When you dump raw bytes in the manner you describe you short-circuit one of the main advantages of using HDF5/PyTables. If you do short-circuit, you have a high probability that the resulting file will look like garbage on anything but the original system that produced it.
So in summary, you should be converting the chunks to the appropriate data types in memory and then writing out. Yes this takes more processing power, time, etc. So in addition to being the right thing to do it will ultimately save you huge headaches down the road.
Related
I am trying to do apriori association mining with WEKA (i use 3.7) using given database table
So, i exported two columns (orderLineNumber and productCode) and load it into weka, as far as i go, i haven't got any success attempt, always ended with "No large itemsets and rules found!"
Again, i tried to convert the csv into ARFF file first using ARFF Converter and still get the same message;
I also tried using database loader in WEKA, the data loaded just fine but still give the same result;
The filter i've applied in preprocessing is only numericToNominal filter;
What have i wrongly done here, i suspiciously think it was my ARFF format though, thank you
Update
After further trial, i found out that i exported wrong column and i lack 1 filter process, which is "denormalized", i installed the plugin via packet manager and denormalized my data after converting it to nominal first;
I then compared the results with "Supermarket" sample's result; The only difference are my output came with 'f' instead of 't' (like shown below) and the confidence value seems like always 100%;
First of all, OrderLine is the wrong column.
Obviously, the position on the printed bill is not very important.
Secondly, the file format is not appropriate.
You want one line for every order, one column for every possible item in the #data section. To save memory, it may be helpful to use sparse formats (do not forget to set flags appropriately)
Other tools like ELKI can process input formats like this, that may be easier to use (it also was a lot faster than Weka):
apple banana
milk diapers beer
but last I checked, ELKI would "only" find frequent itemsets (the harder part) not compute association rules. I then used a tiny python script to produce actual association rules as desired.
I work for a research consortium with a web-based data management system that's managed by another agency. I can download the underlying data from that system as a collection of CSV files. Using R and knitr, I've built a moderately complex reporting system on top of those files. But every so often, the other agency changes the format of the data extracts and blows up my reports (or worse, changes it in a subtle yet nefarious way that I don't notice for weeks).
They'll probably never notify me when these things happen, so I suppose I should be testing more. I'd like to start by testing that those CSV files have the same structure each time (but allowing different numbers of rows as we collect more data). What's the best way to do that? R is my preferred tool but I'm interested in hearing about others that are free and on Windows.
If your files are just CSVs, here is an example (assuming you keep a reference file around):
reference.file <- read.csv("ref.csv")
new.file <- read.csv("new.file")
struct.extract <- function(df) {
list(
vapply(df, class, character(1L)),
attributes(df)[names(attributes(df)) != "row.names"]
)
}
identical(struct.extract(reference.file), struct.extract(new.file))
This compares attributes of the data frame, as well as classes of the columns. If you need to get more detailed on column format you can extend this easily. This assumes the reports are not changing # of rows (or columns), but if that's the case, that should be easy to modify as well.
I have a moderate-sized file (4GB CSV) on a computer that doesn't have sufficient RAM to read it in (8GB on 64-bit Windows). In the past I would just have loaded it up on a cluster node and read it in, but my new cluster seems to arbitrarily limit processes to 4GB of RAM (despite the hardware having 16GB per machine), so I need a short-term fix.
Is there a way to read in part of a CSV file into R to fit available memory limitations? That way I could read in a third of the file at a time, subset it down to the rows and columns I need, and then read in the next third?
Thanks to commenters for pointing out that I can potentially read in the whole file using some big memory tricks:
Quickly reading very large tables as dataframes in R
I can think of some other workarounds (e.g. open in a good text editor, lop off 2/3 of the observations, then load in R), but I'd rather avoid them if possible.
So reading it in pieces still seems like the best way to go for now.
After reviewing this thread I noticed a conspicuous solution to this problem was not mentioned. Use connections!
1) Open a connection to your file
con = file("file.csv", "r")
2) Read in chunks of code with read.csv
read.csv(con, nrows="CHUNK SIZE",...)
Side note: defining colClasses will greatly speed things up. Make sure to define unwanted columns as NULL.
3) Do what ever you need to do
4) Repeat.
5) Close the connection
close(con)
The advantage of this approach is connections. If you omit this step, it will likely slow things down a bit. By opening a connection manually, you essentially open the data set and do not close it until you call the close function. This means that as you loop through the data set you will never lose your place. Imagine that you have a data set with 1e7 rows. Also imagine that you want to load a chunk of 1e5 rows at a time. Since we open the connection we get the first 1e5 rows by running read.csv(con, nrow=1e5,...), then to get the second chunk we run read.csv(con, nrow=1e5,...) as well, and so on....
If we did not use the connections we would get the first chunk the same way, read.csv("file.csv", nrow=1e5,...), however for the next chunk we would need to read.csv("file.csv", skip = 1e5, nrow=2e5,...). Clearly this is inefficient. We are have to find the 1e5+1 row all over again, despite the fact that we just read in the 1e5 row.
Finally, data.table::fread is great. But you can not pass it connections. So this approach does not work.
I hope this helps someone.
UPDATE
People keep upvoting this post so I thought I would add one more brief thought. The new readr::read_csv, like read.csv, can be passed connections. However, it is advertised as being roughly 10x faster.
You could read it into a database using RSQLite, say, and then use an sql statement to get a portion.
If you need only a single portion then read.csv.sql in the sqldf package will read the data into an sqlite database. First, it creates the database for you and the data does not go through R so limitations of R won't apply (which is primarily RAM in this scenario). Second, after loading the data into the database , sqldf reads the output of a specified sql statement into R and finally destroys the database. Depending on how fast it works with your data you might be able to just repeat the whole process for each portion if you have several.
Only one line of code accomplishes all three steps, so it's a no-brainer to just try it.
DF <- read.csv.sql("myfile.csv", sql=..., ...other args...)
See ?read.csv.sql and ?sqldf and also the sqldf home page.
Reading ~5x10^6 numeric values into R from a text file is relatively slow on my machine (a few seconds, and I read several such files), even with scan(..., what="numeric", nmax=5000) or similar tricks. Could it be worthwhile to try an Rcpp wrapper for this sort of task (e.g. Armadillo has a few utilities to read text files)?
Or would I likely be wasting my time for little to no gain in performance because of an expected interface overhead? I'm not sure what's currently limiting the speed (intrinsic machine performance, or else?) It's a task that I repeat many times a day, typically, and the file format is always the same, 1000 columns, around 5000 rows.
Here's a sample file to play with, if needed.
nr <- 5000
nc <- 1000
m <- matrix(round(rnorm(nr*nc),3),nr=nr)
cat(m[1, -1], "\n", file = "test.txt") # first line is shorter
write.table(m[-1, ], file = "test.txt", append=TRUE,
row.names = FALSE, col.names = FALSE)
Update: I tried read.csv.sql and also load("test.txt", arma::raw_ascii) using Armadillo and both were slower than the scan solution.
I highly recommend checking out fread in the latest version of data.table. The version on CRAN (1.8.6) doesn't have fread yet (at the time of this post) so you should be able to get it if you install from the latest source at R-forge. See here.
Please bear in mind that I'm not an R-expert but maybe the concept applies here too: usually reading binary stuff is much faster than reading text files. If your source files don't change frequently (e.g. you are running varied versions of your script/program on the same data), try to read them via scan() once and store them in a binary format (the manual has a chapter about exporting binary files).
From there on you can modify your program to read the binary input.
#Rcpp: scan() & friends are likely to call a native implementation (like fscanf()) so writing your own file read functions via Rcpp may not provide a huge performance gain. You can still try it though (and optimize for your particular data).
Salut Baptiste,
Data Input/Output is a huge topic, so big that R comes with its own manual on data input/output.
R's basic functions can be slow because they are so very generic. If you know your format, you can easily write yourself a faster import adapter. If you know your dimensions too, it is even easier as you need only one memory allocation.
Edit: As a first approximation, I would write a C++ ten-liner. Open a file, read a line, break it into tokens, assign to a vector<vector< double > > or something like that. Even if you use push_back() on individual vector elements, you should be competitive with scan(), methinks.
I once had a neat little csv reader class in C++ based on code by Brian Kernighan himself. Fairly generic (for csv files), fairly powerful.
You can then squeeze performance as you see fit.
Further edit: This SO question has a number of pointers for the csv reading case, and references to the Kernighan and Plauger book.
Yes, you almost certainly can create something that goes faster than read.csv/scan. However, for high performance file reading there are some existing tricks that already let you go much faster, so anything you do would be competing against those.
As Mathias alluded to, if your files don't change very often, then you can cache them by calling save, then restore them with load. (Make sure to use ascii = FALSE, since reading the binary files will be quicker.)
Secondly, as Gabor mentioned, you can often get a substantial performance boost by reading your file into a database and then from that database into R.
Thirdly, you can use the HadoopStreaming package to use Hadoop's file reading capabilities.
For more thoughts in these techniques, see Quickly reading very large tables as dataframes in R.
I'm wondering if there's any documentation about the efficiency of operations in R, specifically those related to data manipulation.
For example:
I imagine it's efficient to add columns to a data frame, because I'm guessing you're just adding an element to a linked list.
I imagine adding rows is slower because vectors are held in arrays at the C level and you have to allocate a new array of length n+1 and copy all the elements over.
The developers probably don't want to tie themselves to a particular implementation, but it would be nice to have something more solid than guesses to go on.
Also, I know the main R performance hint is to use vectored operations whenever possible as opposed to loops.
what about the various flavors of apply?
are those just hidden loops?
what about matrices vs. data frames?
Data IO was one of the features i looked into before i committed to learning R. For better or worse, here are my observations and solutions/palliatives on these issues:
1. That R doesn't handle big data (>2 GB?) To me this is a misnomer. By default, the common data input functions load your data into RAM. Not to be glib, but to me, this is a feature not a bug--anytime my data will fit in my available RAM, that's where i want it. Likewise, one of SQLite's most popular features is the in-memory option--the user has the easy option of loading the entire dB into RAM. If your data won't fit in memory, then R makes it astonishingly easy to persist it, via connections to the common RDBMS systems (RODBC, RSQLite, RMySQL, etc.), via no-frills options like the filehash package, and via systems that current technology/practices (for instance, i can recommend ff). In other words, the R developers have chosen a sensible (and probably optimal) default, from which it is very easy to opt out.
2. The performance of read.table (read.csv, read.delim, et al.), the most common means for getting data into R, can be improved 5x (and often much more in my experience) just by opting out of a few of read.table's default arguments--the ones having the greatest effect on performance are mentioned in the R's Help (?read.table). Briefly, the R Developers tell us that if you provide values for the parameters 'colClasses', 'nrows', 'sep', and 'comment.char' (in particular, pass in '' if you know your file begins with headers or data on line 1), you'll see a significant performance gain. I've found that to be true.
Here are the snippets i use for those parameters:
To get the number of rows in your data file (supply this snippet as an argument to the parameter, 'nrows', in your call to read.table):
as.numeric((gsub("[^0-9]+", "", system(paste("wc -l ", file_name, sep=""), intern=T))))
To get the classes for each column:
function(fname){sapply(read.table(fname, header=T, nrows=5), class)}
Note: You can't pass this snippet in as an argument, you have to call it first, then pass in the value returned--in other words, call the function, bind the returned value to a variable, and then pass in the variable as the value to to the parameter 'colClasses' in your call to read.table:
3. Using Scan. With only a little more hassle, you can do better than that (optimizing 'read.table') by using 'scan' instead of 'read.table' ('read.table' is actually just a wrapper around 'scan'). Once again, this is very easy to do. I use 'scan' to input each column individually then build my data.frame inside R, i.e., df = data.frame(cbind(col1, col2,....)).
4. Use R's Containers for persistence in place of ordinary file formats (e.g., 'txt', 'csv'). R's native data file '.RData' is a binary format that a little smaller than a compressed ('.gz') txt data file. You create them using save(, ). You load it back into the R namespace with load(). The difference in load times compared with 'read.table' is dramatic. For instance, w/ a 25 MB file (uncompressed size)
system.time(read.table("tdata01.txt.gz", sep=","))
=> user system elapsed
6.173 0.245 **6.450**
system.time(load("tdata01.RData"))
=> user system elapsed
0.912 0.006 **0.912**
5. Paying attention to data types can often give you a performance boost and reduce your memory footprint. This point is probably more useful in getting data out of R. The key point to keep in mind here is that by default, numbers in R expressions are interpreted as double-precision floating point, e.g., > typeof(5) returns "double." Compare the object size of a reasonable-sized array of each and you can see the significance (use object.size()). So coerce to integer when you can.
Finally, the 'apply' family of functions (among others) are not "hidden loops" or loop wrappers. They are loops implemented in C--big difference performance-wise. [edit: AWB has correctly pointed out that while 'sapply', 'tapply', and 'mapply' are implemented in C, 'apply' is simply a wrapper function.
These things do pop up on the lists, in particular on r-devel. One fairly well-established nugget is that e.g. matrix operations tend to be faster than data.frame operations. Then there are add-on packages that do well -- Matt's data.table package is pretty fast, and Jeff has gotten xts indexing to be quick.
But it "all depends" -- so you are usually best adviced to profile on your particular code. R has plenty of profiling support, so you should use it. My Intro to HPC with R tutorials have a number of profiling examples.
I will try to come back and provide more detail. If you have any question about the efficiency of one operation over another, you would do best to profile your own code (as Dirk suggests). The system.time() function is the easiest way to do this although there are many more advanced utilities (e.g. Rprof, as documented here).
A quick response for the second part of your question:
What about the various flavors of apply? Are those just hidden loops?
For the most part yes, the apply functions are just loops and can be slower than for statements. Their chief benefit is clearer code. The main exception that I have found is lapply which can be faster because it is coded in C directly.
And what about matrices vs. data frames?
Matrices are more efficient than data frames because they require less memory for storage. This is because data frames require additional attribute data. From R Introduction:
A data frame may for many purposes be regarded as a matrix with columns possibly of differing modes and attributes