Related
I have a #transform_pandas code which loads the input file for computing.
Inside the compute function I have a for loop which has to read the complete input data and filter accordingly for every iteration.
#transform_pandas(
Output("/FCA_Foundry/dataset1"),
source_df=Input(sample),
)
I have the below code where I'm trying to read source_df dataset for every iteration in for loop and filter the dataset specifically to the year and family and do the computation.
def compute(source_df):
for entire_row in vhcl_df.itertuples():
modyr = entire_row[1]
fam = str(entire_row[2])
/* source_df should be read again here.
source_df = source_df.loc[source_df['i_yr']==modyr]
source_df = source_df.loc[source_df['fam']==fam]
...
Is there a way to achieve this. Thank you for your support.
As already suggested by #nicornk in the comments, you should create a new .copy() item of your source_df right after you declare the transform.
The two filtering steps (that can ben also merged in one, if you don't need to work just on the "modyr filtered" source_df.
Please note that modyr, fam are actual colnames of vhcl_df, it is actually sufficient to
#transform_pandas(
Output("/FCA_Foundry/dataset1"),
source_df=Input(sample),
vhcl_df=Input(path)
)
def compute(source_df, vhcl_df):
for modyr, fam in vhcl_df.items():
temp_df = source_df.copy()
temp_df = source_df.loc[source_df['i_yr']==modyr]
temp_df = source_df.loc[source_df['fam']==str(fam)]
which, in a more concise and clean way is actually writable as
def compute(source_df, vhcl_df):
for modyr, fam in vhcl_df.items():
temp_df = source_df.copy()
filtered_temp_df = temp_df[(temp_df.i_yr==modyr) & (temp_df.fam==str(fam))]
PS: Remember that if source_df is big, you should proceed with PySpark (see foundry docs)
Note that transform_pandas should only be used on datasets that can fit into memory. If you have larger datasets that you wish to filter down first before converting to Pandas, you should write your transformation using the transform_df() decorator and the pyspark.sql.SparkSession.createDataFrame() method.
This sounds like it should be REALLY easy to answer with Google but I'm finding it impossible to answer the majority of my nontrivial pandas/pytables questions this way. All I'm trying to do is to load about 3 billion records from about 6000 different CSV files into a single table in a single HDF5 file. It's a simple table, 26 fields, mixture of strings, floats and ints. I'm loading the CSVs with df = pandas.read_csv() and appending them to my hdf5 file with df.to_hdf(). I really don't want to use df.to_hdf(data_columns = True) because it looks like that will take about 20 days versus about 4 days for df.to_hdf(data_columns = False). But apparently when you use df.to_hdf(data_columns = False) you end up with some pile of junk that you can't even recover the table structure from (or so it appears to my uneducated eye). Only the columns that were identified in the min_itemsize list (the 4 string columns) are identifiable in the hdf5 table, the rest are being dumped by data type into values_block_0 through values_block_4:
table = h5file.get_node('/tbl_main/table')
print(table.colnames)
['index', 'values_block_0', 'values_block_1', 'values_block_2', 'values_block_3', 'values_block_4', 'str_col1', 'str_col2', 'str_col3', 'str_col4']
And any query like df = pd.DataFrame.from_records(table.read_where(condition)) fails with error "Exception: Data must be 1-dimensional"
So my questions are: (1) Do I really have to use data_columns = True which takes 5x as long? I was expecting to do a fast load and then index just a few columns after loading the table. (2) What exactly is this pile of garbage I get using data_columns = False? Is it good for anything if I need my table back with query-able columns? Is it good for anything at all?
This is how you can create an HDF5 file from CSV data using pytables. You could also use a similar process to create the HDF5 file with h5py.
Use a loop to read the CSV files with np.genfromtxt into a np array.
After reading the first CSV file, write the data with .create_table() method, referencing the np array created in Step 1.
For additional CSV files, write the data with .append() method, referencing the np array created in Step 1
End of loop
Updated on 6/2/2019 to read a date field (mm/dd/YYY) and convert to datetime object. Note changes to genfromtxt() arguments! Data used is added below the updated code.
import numpy as np
import tables as tb
from datetime import datetime
csv_list = ['SO_56387241_1.csv', 'SO_56387241_2.csv' ]
my_dtype= np.dtype([ ('a',int),('b','S20'),('c',float),('d',float),('e','S20') ])
with tb.open_file('SO_56387241.h5', mode='w') as h5f:
for PATH_csv in csv_list:
csv_data = np.genfromtxt(PATH_csv, names=True, dtype=my_dtype, delimiter=',', encoding=None)
# modify date in fifth field 'e'
for row in csv_data :
datetime_object = datetime.strptime(row['my_date'].decode('UTF-8'), '%m/%d/%Y' )
row['my_date'] = datetime_object
if h5f.__contains__('/CSV_Data') :
dset = h5f.root.CSV_Data
dset.append(csv_data)
else:
dset = h5f.create_table('/','CSV_Data', obj=csv_data)
dset.flush()
h5f.close()
Data for testing:
SO_56387241_1.csv:
my_int,my_str,my_float,my_exp,my_date
0,zero,0.0,0.00E+00,01/01/1980
1,one,1.0,1.00E+00,02/01/1981
2,two,2.0,2.00E+00,03/01/1982
3,three,3.0,3.00E+00,04/01/1983
4,four,4.0,4.00E+00,05/01/1984
5,five,5.0,5.00E+00,06/01/1985
6,six,6.0,6.00E+00,07/01/1986
7,seven,7.0,7.00E+00,08/01/1987
8,eight,8.0,8.00E+00,09/01/1988
9,nine,9.0,9.00E+00,10/01/1989
SO_56387241_2.csv:
my_int,my_str,my_float,my_exp,my_date
10,ten,10.0,1.00E+01,01/01/1990
11,eleven,11.0,1.10E+01,02/01/1991
12,twelve,12.0,1.20E+01,03/01/1992
13,thirteen,13.0,1.30E+01,04/01/1993
14,fourteen,14.0,1.40E+01,04/01/1994
15,fifteen,15.0,1.50E+01,06/01/1995
16,sixteen,16.0,1.60E+01,07/01/1996
17,seventeen,17.0,1.70E+01,08/01/1997
18,eighteen,18.0,1.80E+01,09/01/1998
19,nineteen,19.0,1.90E+01,10/01/1999
I am working on a qualitative analysis project in the tm package of R. I have built a corpus and created a term document matrix and long story short I need to edit my term document matrix and conflate some of its rows. To do this I have exported it out of R using
write.csv()
I then have imported the csv file back into R but am struggling to figure out how to get R to read it as a TermDocumentMatrix or DocumentTermMatrix.
I tried using the suggestions of the following example code with no avail.
It seems to keep reading my matrix as if it was a corpus and each cell as a single document.
# change this file location to suit your machine
file_loc <- "C:\\Documents and Settings\\Administrator\\Desktop\\Book1.csv"
# change TRUE to FALSE if you have no column headings in the CSV
x <- read.csv(file_loc, header = TRUE)
require(tm)
corp <- Corpus(DataframeSource(x))
dtm <- DocumentTermMatrix(corp)
Is there any way to import in a csv matrix that will be read as a termdocumentmatrix or documenttermmatrix without having R read the csv as if each cell is a document?
You're not reading documents, so skip the Corpus() step. This should work directly:
myDTM <- as.DocumentTermMatrix(x, weighting = weightTf)
For next time, consider saving the TDM object as .RData as this will not require conversion, and is also much more efficient.
If you want to keep the format of any data, I would recommend to use the save() function.
You can save any R objects into a .RData file. And when you want to retrieve the data, you can use the load() function.
For my dissertation data collection, one of the sources is an externally-managed system, which is based on Web form for submitting SQL queries. Using R and RCurl, I have implemented an automated data collection framework, where I simulate the above-mentioned form. Everything worked well while I was limiting the size of the resulting dataset. But, when I tried to go over 100000 records (RQ_SIZE in the code below), the tandem "my code - their system" started being unresponsive ("hanging").
So, I have decided to use SQL pagination feature (LIMIT ... OFFSET ...) to submit a series of requests, hoping then to combine the paginated results into a target data frame. However, after changing my code accordingly, the output that I see is only one pagination progress character (*) and then no more output. I'd appreciate, if you could help me identify the probable cause of the unexpected behavior. I cannot provide reproducible example, as it's very difficult to extract the functionality, not to mention the data, but I hope that the following code snippet would be enough to reveal the issue (or, at least, a direction toward the problem).
# First, retrieve total number of rows for the request
srdaRequestData(queryURL, "COUNT(*)", rq$from, rq$where,
DATA_SEP, ADD_SQL)
assign(dataName, srdaGetData()) # retrieve result
data <- get(dataName)
numRequests <- as.numeric(data) %/% RQ_SIZE + 1
# Now, we can request & retrieve data via SQL pagination
for (i in 1:numRequests) {
# setup SQL pagination
if (rq$where == '') rq$where <- '1=1'
rq$where <- paste(rq$where, 'LIMIT', RQ_SIZE, 'OFFSET', RQ_SIZE*(i-1))
# Submit data request
srdaRequestData(queryURL, rq$select, rq$from, rq$where,
DATA_SEP, ADD_SQL)
assign(dataName, srdaGetData()) # retrieve result
data <- get(dataName)
# some code
# add current data frame to the list
dfList <- c(dfList, data)
if (DEBUG) message("*", appendLF = FALSE)
}
# merge all the result pages' data frames
data <- do.call("rbind", dfList)
# save current data frame to RDS file
saveRDS(data, rdataFile)
It probably falls into the category when presumably MySQL hinders LIMIT OFFSET:
Why does MYSQL higher LIMIT offset slow the query down?
Overall, fetching large data sets over HTTP repeatedly is not very reliable.
Since this is for your dissertation, here is a hand:
## Folder were to save the results to disk.
## Ideally, use a new, empty folder. Easier then to load from disk
folder.out <- "~/mydissertation/sql_data_scrape/"
## Create the folder if not exist.
dir.create(folder.out, showWarnings=FALSE, recursive=TRUE)
## The larger this number, the more memory you will require.
## If you are renting a large box on, say, EC2, then you can make this 100, or so
NumberOfOffsetsBetweenSaves <- 10
## The limit size per request
RQ_SIZE <- 1000
# First, retrieve total number of rows for the request
srdaRequestData(queryURL, "COUNT(*)", rq$from, rq$where,
DATA_SEP, ADD_SQL)
## Get the total number of rows
TotalRows <- as.numeric(srdaGetData())
TotalNumberOfRequests <- TotalRows %/% RQ_SIZE
TotalNumberOfGroups <- TotalNumberOfRequests %/% NumberOfOffsetsBetweenSaves + 1
## FYI: Total number of rows being requested is
## (NumberOfOffsetsBetweenSaves * RQ_SIZE * TotalNumberOfGroups)
for (g in seq(TotalNumberOfGroups)) {
ret <-
lapply(seq(NumberOfOffsetsBetweenSaves), function(i) {
## function(i) is the same code you have
## inside your for loop, but cleaned up.
# setup SQL pagination
if (rq$where == '')
rq$where <- '1=1'
rq$where <- paste(rq$where, 'LIMIT', RQ_SIZE, 'OFFSET', RQ_SIZE*g*(i-1))
# Submit data request
srdaRequestData(queryURL, rq$select, rq$from, rq$where,
DATA_SEP, ADD_SQL)
# retrieve result
data <- srdaGetData()
# some code
if (DEBUG) message("*", appendLF = FALSE)
### DONT ASSIGN TO dfList, JUST RETURN `data`
# xxxxxx DONT DO: xxxxx dfList <- c(dfList, data)
### INSTEAD:
## return
data
})
## save each iteration
file.out <- sprintf("%s/data_scrape_%04i.RDS", folder.out, g)
saveRDS(do.call(rbind, ret), file=file.out)
## OPTIONAL (this will be slower, but will keep your rams and goats in line)
# rm(ret)
# gc()
}
Then, once you are done scraping:
library(data.table)
folder.out <- "~/mydissertation/sql_data_scrape/"
files <- dir(folder.out, full=TRUE, pattern="\\.RDS$")
## Create an empty list
myData <- vector("list", length=length(files))
## Option 1, using data.frame
for (i in seq(myData))
myData[[i]] <- readRDS(files[[i]])
DT <- do.call(rbind, myData)
## Option 2, using data.table
for (i in seq(myData))
myData[[i]] <- as.data.table(readRDS(files[[i]]))
DT <- rbindlist(myData)
I'm answering my own question, as, finally, I have figured out what has been the real source of the problem. My investigation revealed that the unexpected waiting state of the program was due to PostgreSQL becoming confused by malformed SQL queries, which contained multiple LIMIT and OFFSET keywords.
The reason of that is pretty simple: I used rq$where both outside and inside the for loop, which made paste() concatenate previous iteration's WHERE clause with the current one. I have fixed the code by processing contents of the WHERE clause and saving it before the loop and then using the saved value in each iteration of the loop safely, as it became independent from the value of the original WHERE clause.
This investigation also helped me to fix some other deficiencies in my code and make improvements (such as using sub-selects to properly handle SQL queries returning number of records for queries with aggregate functions). The moral of the story: you can never be too careful in software development. Big thank you to those nice people who helped with this question.
I have many "problems" but i will try to split them up as good as i can.
First i will present my code:
# Laster pakker
library(RODBC)
library(plyr)
library(lattice)
library(colorRamps)
library(Perfor)
# Picking up data
query <- "select convert(varchar(10),fk_dim_date,103) fk_dim_date,fk_dim_portfolio, dtd_portfolio_return_pct, dtd_benchmark_return_pct, * from nbim_dm..v_fact_performance
where fk_dim_date > '20130103' and fk_dim_portfolio in ('6906', '1812964')
"
# Formatting SQLen
query <- strwrap(query,width=nchar(query),simplify=TRUE)
# quering
ch <- odbcDriverConnect("driver={SQL Server};server=XXXX;Database=XXXX;", rows_at_time = 1024)
result <- sqlQuery(ch, query, as.is=c(TRUE, TRUE, TRUE))
close(ch)
# Do some cleanup
`enter code here`resultt$v_d <- as.Date(as.POSIXct(t$v_d))
#split
y <- split(qt,qt$fk_dim_portfolio)
#making names
new_names <- c("one","two")
for (i in 1:length(y){assign(new_names[i],y[[i]])})
So far so good:
The table that my SQL is running on has approx 178 diff. port_ids, some of which are useless and others that are highly useful. However i want this code to pull all fk_dim_ports (pulling: '6906', '1812964 was just for example purposes). After pulling the data i want to seperate it into n (now 178 sets) and make them xts objects which i have run into some trouble using:
qt <- xts(t[,-1],order.by=t[,1])
But works perfectly well if i don`t split the data using:
y <- split(qt,qt$fk_dim_portfolio)
Assuming this will work, my intention is to create charts.PerformanceSummary(mydata) for every table of my previous created data frames.
If you have any tips on how to split, make timeseries objects and loop the generation of the charts i would highly appreciate this.
I am aware that this post probably don`t comply to your rules/customs etc, but thanks for helping.
Lars