We use grep, cut, sort, uniq, and join at the command line all the time to do data analysis. They work great, although there are shortcomings. For example, you have to give column numbers to each tool. We often have wide files (many columns) and a column header that gives column names. In fact, our files look a lot like SQL tables. I'm sure there is a driver (ODBC?) that will operate on delimited text files, and some query engine that will use that driver, so we could just use SQL queries on our text files. Since doing analysis is usually ad hoc, it would have to be minimal setup to query new files (just use the files I specify in this directory) rather than declaring particular tables in some config.
Practically speaking, what's the easiest? That is, the SQL engine and driver that is easiest to set up and use to apply against text files?
David Malcolm wrote a little tool named "squeal" (formerly "show"), which allows you to use SQL-like command-line syntax to parse text files of various formats, including CSV.
An example on squeal's home page:
$ squeal "count(*)", source from /var/log/messages* group by source order by "count(*)" desc
count(*)|source |
--------+--------------------+
1633 |kernel |
1324 |NetworkManager |
98 |ntpd |
70 |avahi-daemon |
63 |dhclient |
48 |setroubleshoot |
39 |dnsmasq |
29 |nm-system-settings |
27 |bluetoothd |
14 |/usr/sbin/gpm |
13 |acpid |
10 |init |
9 |pcscd |
9 |pulseaudio |
6 |gnome-keyring-ask |
6 |gnome-keyring-daemon|
6 |gnome-session |
6 |rsyslogd |
5 |rpc.statd |
4 |vpnc |
3 |gdm-session-worker |
2 |auditd |
2 |console-kit-daemon |
2 |libvirtd |
2 |rpcbind |
1 |nm-dispatcher.action|
1 |restorecond |
q - Run SQL directly on CSV or TSV files:
https://github.com/harelba/q
Riffing off someone else's suggestion, here is a Python script for sqlite3. A little verbose, but it works.
I don't like having to completely copy the file to drop the header line, but I don't know how else to convince sqlite3's .import to skip it. I could create INSERT statements, but that seems just as bad if not worse.
Sample invocation:
$ sql.py --file foo --sql "select count(*) from data"
The code:
#!/usr/bin/env python
"""Run a SQL statement on a text file"""
import os
import sys
import getopt
import tempfile
import re
class Usage(Exception):
def __init__(self, msg):
self.msg = msg
def runCmd(cmd):
if os.system(cmd):
print "Error running " + cmd
sys.exit(1)
# TODO(dan): Return actual exit code
def usage():
print >>sys.stderr, "Usage: sql.py --file file --sql sql"
def main(argv=None):
if argv is None:
argv = sys.argv
try:
try:
opts, args = getopt.getopt(argv[1:], "h",
["help", "file=", "sql="])
except getopt.error, msg:
raise Usage(msg)
except Usage, err:
print >>sys.stderr, err.msg
print >>sys.stderr, "for help use --help"
return 2
filename = None
sql = None
for o, a in opts:
if o in ("-h", "--help"):
usage()
return 0
elif o in ("--file"):
filename = a
elif o in ("--sql"):
sql = a
else:
print "Found unexpected option " + o
if not filename:
print >>sys.stderr, "Must give --file"
sys.exit(1)
if not sql:
print >>sys.stderr, "Must give --sql"
sys.exit(1)
# Get the first line of the file to make a CREATE statement
#
# Copy the rest of the lines into a new file (datafile) so that
# sqlite3 can import data without header. If sqlite3 could skip
# the first line with .import, this copy would be unnecessary.
foo = open(filename)
datafile = tempfile.NamedTemporaryFile()
first = True
for line in foo.readlines():
if first:
headers = line.rstrip().split()
first = False
else:
print >>datafile, line,
datafile.flush()
#print datafile.name
#runCmd("cat %s" % datafile.name)
# Create columns with NUMERIC affinity so that if they are numbers,
# SQL queries will treat them as such.
create_statement = "CREATE TABLE data (" + ",".join(
map(lambda x: "`%s` NUMERIC" % x, headers)) + ");"
cmdfile = tempfile.NamedTemporaryFile()
#print cmdfile.name
print >>cmdfile,create_statement
print >>cmdfile,".separator ' '"
print >>cmdfile,".import '" + datafile.name + "' data"
print >>cmdfile, sql + ";"
cmdfile.flush()
#runCmd("cat %s" % cmdfile.name)
runCmd("cat %s | sqlite3" % cmdfile.name)
if __name__ == "__main__":
sys.exit(main())
Maybe write a script that creates an SQLite instance (possibly in memory), imports your data from a file/stdin (accepting your data's format), runs a query, then exits?
Depending on the amount of data, performance could be acceptable.
MySQL has a CVS storage engine, that might do what you need, if your files are CSV files.
Otherwise, you can use mysqlimport to import text files into MySQL. You could create a wrapper around mysqlimport, which figures out columns etc. and creates the necessary table.
You might also be able to use DBD::AnyData, a Perl module which lets you access text files like a database.
That said, it sounds a lot like you should really look at using a database. Is it really easier keeping table-oriented data in text files?
I have used Microsoft LogParser to query csv files several times... and it serves the purpose. It was surprising to see such a useful tool from M$ that too Free!
Related
Whenever I use read.csv.sql I cannot select from the first column with and any output from the code places an unusual character (A(tilde)-..) at the begging of the first column's name.
So suppose I create a df.csv file in in Excel that looks something like this
df = data.frame(
a = 1,
b = 2,
c = 3,
d = 4)
Then if I use sqldf to query the csv which is in my working directory I get the following error:
> read.csv.sql("df.csv", sql = "select * from file where a == 1")
Error in result_create(conn#ptr, statement) : no such column: a
If I query a different column than the first, I get a result but with the output of the unusual characters as seen below
df <- read.csv.sql("df.csv", sql = "select * from file where b == 2")
View(df)
Any idea how to prevent these characters from being added to the first column name?
The problem is presumably that you have a file that is larger than R can handle and so only want to read a subset of rows into R and specifying the condition to filter it by involves referring to the first column whose name is messed up so you can't use it.
Here are two alternative approaches. The first one involves a bit more code but has the advantage that it is 100% R. The second one is only one statement and also uses R but additionally makes use an of an external utility.
1) skip header Read the file in skipping over the header. That will cause the columns to be labelled V1, V2, etc. and use V1 in the condition.
# write out a test file - BOD is a data frame that comes with R
write.csv(BOD, "BOD.csv", row.names = FALSE, quote = FALSE)
# read file skipping over header
DF <- read.csv.sql("BOD.csv", "select * from file where V1 < 3",
skip = 1, header = FALSE)
# read in header, assign it to DF and fix first column
hdr <- read.csv.sql("BOD.csv", "select * from file limit 0")
names(DF) <- names(hdr)
names(DF)[1] <- "TIME" # suppose we want TIME instead of Time
DF
## TIME demand
## 1 1 8.3
## 2 2 10.3
2) filter Another way to proceed is to use the filter= argument. Here we assume we know that the end of the column name is ime but there are other characters prior to that that we don't know. This assumes that sed is available and on your path. If you are on Windows install Rtools to get sed. The quoting might need to be changed depending on your shell.
When trying this on Windows I noticed that sed from Rtools changed the line endings so below we specified eol= to ensure correct processing. You may not need that.
DF <- read.csv.sql("BOD.csv", "select * from file where TIME < 3",
filter = 'sed -e "1s/.*ime,/TIME,/"' , eol = "\n")
DF
## TIME demand
## 1 1 8.3
## 2 2 10.3
So I figured it out by reading through the above comments.
I'm on a Windows 10 machine using Excel for Office 365. The special characters will go away by changing how I saved the file from a "CSV UTF-8 (Comma Delimited)" to just "CSV (Comma delimited)".
I am using read.csv.sql to conditionally read in data (my data set is extremely large so this was the solution I chose to filter it and reduce it in size prior to reading the data in). I was running into memory issues by reading in the full data and then filtering it so that is why it is important that I use the conditional read so that the subset is read in, versus the full data set.
Here is a small data set so my problem can be reproduced:
write.csv(iris, "iris.csv", row.names = F)
library(sqldf)
csvFile <- "iris.csv"
I am finding that the notation you have to use is extremely awkward using read.csv.sql the following is the how I am reading in the file:
# Step 1 (Assume these values are coming from UI)
spec <- 'setosa'
petwd <- 0.2
# Add quotes and make comma-separated:
spec <- toString(sprintf("'%s'", spec))
petwd <- toString(sprintf("'%s'", petwd))
# Step 2 - Conditionally read in the data, store in 'd'
d <- fn$read.csv.sql(csvFile, sql='select * from file where
"Species" in ($spec)'
and "Petal.Width" in ($petwd)',
filter = list('gawk -f prog', prog = '{ gsub(/"/, ""); print }'))
My main problem is that if any of the values above (from UI) are null then it won't read in the data properly, because this chunk of code is all hard coded.
I would like to change this into: Step 1 - check which values are null and do not filter off of them, then filter using read.csv.sql for all non-null values on corresponding columns.
Note: I am reusing the code from this similar question within this question.
UPDATE
I want to clear up what I am asking. This is what I am trying to do:
If a field, say spec comes through as NA (meaning the user did not pick input) then I want it to filter as such (default to spec == EVERY SPEC):
# Step 2 - Conditionally read in the data, store in 'd'
d <- fn$read.csv.sql(csvFile, sql='select * from file where
"Petal.Width" in ($petwd)',
filter = list('gawk -f prog', prog = '{ gsub(/"/, ""); print }'))
Since spec is NA, if you try to filter/read in a file matching spec == NA it will read in an empty data set since there are no NA values in my data, hence breaking the code and program. Hope this clears it up more.
There are several problems:
some of the simplifications provided in the link in the question were not followed.
spec is a scalar so one can just use '$spec'
petwd is a numeric scalar and SQL does not require quotes around numbers so just use $petwd
the question states you want to handle empty fields but not how so we have used csvfix to map them to -1 and also strip off quotes. (Alternately let them enter and do it in R. Empty numerics will come through as 0 and empty character fields will come through as zero length character fields.)
you can use [...] in place of "..." in SQL
The code below worked for me in both Windows and Ubuntu Linux with the bash shell.
library(sqldf)
spec <- 'setosa'
petwd <- 0.2
d <- fn$read.csv.sql(
"iris.csv",
sql = "select * from file where [Species] = '$spec' and [Petal.Width] = $petwd",
verbose = TRUE,
filter = 'csvfix map -smq -fv "" -tv -1'
)
Update
Regarding the update at the end of the question it was clarified that the NA could be in spec as opposed to being in the data being read in and that if spec is NA then the condition involving spec should be regarded as TRUE. In that case just expand the SQL where condition to handle that as follows.
spec <- NA
petwd <- 0.2
d <- fn$read.csv.sql(
"iris.csv",
sql = "select * from file
where ('$spec' == 'NA' or [Species] = '$spec') and [Petal.Width] = $petwd",
verbose = TRUE,
filter = 'csvfix echo -smq'
)
The above will return all rows for which Petal.Width is 0.2 .
I have a very wide Excel sheet, from Column A - DIE (about 2500 columns wide), of survey data. Each column is a question, and each row is a response. I'm trying to upload the data to SQL and convert it to a more SQL-friendly format using the UNPIVOT function, but I can't even get it loaded into SQL because it exceeds the 1024-column limit.
Basically, I have an Excel sheet that looks like this:
But I want to convert it to look like this:
What options do I have to make this change, either in Excel (prior to upload) or SQL (while circumventing the 1024 column limit)?
I have had to do this quite a bit. My solution was to write a Python script that would un-crosstab a CSV file (typically exported from Excel), creating another CSV file. The Python code is here: https://pypi.python.org/pypi/un-xtab/ and the documentation is here: http://pythonhosted.org/un-xtab/. I've never run it on a file with 2500 columns, but don't know why it wouldn't work.
R has a very specific function call in one of it's libraries. You can also connect, read, and write data with R into a database. Would suggest downloading R and Rstudio.
Here is a working script to get you started that does what you need:
Sample data:
df <- data.frame(id = c(1,2,3), question_1 = c(1,0,1), question_2 = c(2,0,2))
df
Input table:
id question_1 question_2
1 1 1 2
2 2 0 0
3 3 1 2
Code to transpose the data:
df2 <- gather(df, key = id, value = values)
df2
Output:
id id values
1 1 question_1 1
2 2 question_1 0
3 3 question_1 1
4 1 question_2 2
5 2 question_2 0
6 3 question_2 2
Some helper functions for you to import and export the csv data:
# Install and load the necessary libraries
install.packages(c('tidyr','readr'))
library(tidyr)
library(readr)
# to read a csv file
df <- read_csv('[some directory][some filename].csv')
# To output the csv file
write.csv(df2, '[some directory]data.csv', row.names = FALSE)
Thanks for all the help. I ended up using Python due to limitations in both SQL (over 1024 columns wide) and Excel (well over 1 million rows in the output). I borrowed the concepts from rd_nielson's code, but that was a bit more complicated than I needed. In case it's helpful to anyone else, this is the code I used. It outputs a csv file with 3 columns and 14 million rows that I can upload to SQL.
import csv
with open('Responses.csv') as f:
reader = csv.reader(f)
headers = next(reader) # capture current field headers
newHeaders = ['ResponseID','Question','Response'] # establish new header names
with open('PythonOut.csv','w') as outputfile:
writer=csv.writer(outputfile, dialect='excel', lineterminator='\n')
writer.writerow(newHeaders) # write new headers to output
QuestionHeaders = headers[1:len(headers)] # Slice the question headers from original header list
for row in reader:
questionCount = 0 # start counter to loop through each question (column) for every response (row)
while questionCount <= len(QuestionHeaders) - 1:
newRow = [row[0], QuestionHeaders[questionCount], row[questionCount + 1]]
writer.writerow(newRow)
questionCount += 1
I have a huge number of .txt files (maybe around 10 millions) each having the same number of rows/colums. They actually are some single channel images and the pixel values are separated with an space. Here's the code I've written to do the work but it's very slow. I wonder if someone can suggest a more optimized/efficient way of doing this:
require 'torch'
f = assert(io.open(txtFilePath, 'r'))
local tempTensor = torch.Tensor(1, 64, 64):fill(0)
local i = 1
for line in f:lines() do
local l = line:split(' ')
for key, val in ipairs(l) do
tempTensor[{1, i, key}] = tonumber(val)
end
i = i + 1
end
f:close()
In brief, change you source files if it is possible.
The only I can suggest is to use binary data instead of txt as a source.
You have got the long-term methods: f:lines(), line:split(' ') and tonumber(val). All of them are using strings as variables.
As I understood, you have got file like this:
0 10 20
11 18 22
....
so, change your source it into binary like this:
<0><18><20><11><18><22> ...
where <18> is a byte in hex form, that is 12 , <20> is 16 , etc.
to read
fid = io.open(sup_filename, "rb")
while true do
local bytes = fid:read(1)
if bytes == nil then break end -- EOF
local st = bytes[0]
print(st)
end
fid:close()
https://www.lua.org/pil/21.2.2.html
It would be dramatically faster.
May be using regular expressions (instead of :split() and lines()) can help to you but I do not think.
Is there a way to plot a function based on values from a text file?
I know how to define a function in gnuplot and then plot it but that is not what I need.
I have a table with constants for functions that are updated regularly. When this update happens I want to be able to run a script that draws a figure with this new curve. Since there are quite few figures to draw I want to automate the procedure.
Here is an example table with constants:
location a b c
1 1 3 4
2
There are two ways I see to solve the problem but I do not know if and how they can be implemented.
I can then use awk to produce the string: f(x)=1(x)**2+3(x)+4, write it to a file and somehow make gnuplot read this new file and plot on a certain x range.
or use awk inside gnuplot something like f(x) = awk /1/ {print "f(x)="$2 etc., or use awk directly in the plot command.
I any case, I'm stuck and have not found a solution to this problem online, do you have any suggestions?
Another possibilty to have a somewhat generic version for this, you can do the following:
Assume, the parameters are stored in a file parameters.dat with the first line containing the variable names and all others the parameter sets, like
location a b c
1 1 3 4
The script file looks like this:
file = 'parameters.dat'
par_names = system('head -1 '.file)
par_cnt = words(par_names)
# which parameter set to choose
par_line_num = 2
# select the respective string
par_line = system(sprintf('head -%d ', par_line_num).file.' | tail -1')
par_string = ''
do for [i=1:par_cnt] {
eval(word(par_names, i).' = '.word(par_line, i))
}
f(x) = a*x**2 + b*x + c
plot f(x) title sprintf('location = %d', location)
This question (gnuplot store one number from data file into variable) had some hints for me in the first answer.
In my case I have a file which contains parameters for a parabola. I have saved the parameters in gnuplot variables. Then I plot the function containing the parameter variables for each timestep.
#!/usr/bin/gnuplot
datafile = "parabola.txt"
set terminal pngcairo size 1000,500
set xrange [-100:100]
set yrange [-100:100]
titletext(timepar, apar, cpar) = sprintf("In timestep %d we have parameter a = %f, parameter c = %f", timepar, apar, cpar)
do for [step=1:400] {
set output sprintf("parabola%04d.png", step)
# read parameters from file, where the first line is the header, thus the +1
a=system("awk '{ if (NR == " . step . "+1) printf \"%f\", $1}' " . datafile)
c=system("awk '{ if (NR == " . step . "+1) printf \"%f\", $2}' " . datafile)
# convert parameters to numeric format
a=a+0.
c=c+0.
set title titletext(step, a, c)
plot c+a*x**2
}
This gives a series of png files called parabola0001.png,
parabola0002.png,
parabola0003.png,
…, each showing a parabola with the parameters read from the file called parabola.txt. The title contains the parameters of the given time step.
For understanding the gnuplot system() function you have to know that:
stuff inside double quotes is not parsed by gnuplot
the dot is for concatenating strings in gnuplot
the double quotes for the awk printf command have to be escaped, to hide them from gnuplot parser
To test this gnuplot script, save it into a file with an arbitrary name, e.g. parabolaplot.gplot and make it executable (chmad a+x parabolaplot.gplot). The parabola.txt file can be created with
awk 'BEGIN {for (i=1; i<=1000; i++) printf "%f\t%f\n", i/200, i/100}' > parabola.txt
awk '/1/ {print "plot "$2"*x**2+"$3"*x+"$4}' | gnuplot -persist
Will select the line and plot it
This was/is another question about how to extract specific values into variables with gnuplot (maybe it would be worth to create a Wiki entry about this topic).
There is no need for using awk, you can do this simply with gnuplot only (hence platform-independent), even with gnuplot 4.6.0 (March 2012).
You can do a stats (check help stats) and assign the values to variables.
Data: SO15007620_Parameters.txt
location a b c
1 1 3 4
2 -1 2 3
3 2 1 -1
Script: (works with gnuplot 4.6.0, March 2012)
### read parameters from separate file into variables
reset
FILE = "SO15007620_Parameters.txt"
myLine = 1 # line index 0-based
stats FILE u (a=$2, b=$3, c=$4) every ::myLine::myLine nooutput
f(x) = a*x**2 + b*x + c
plot f(x) w l lc rgb "red" ti sprintf("f(x) = %gx^2 + %gx + %g", a,b,c)
### end of script
Result: