[datatable-help] New function fread() in v1.8.7

[Matthew Dowle]

Yes autostart is the line it detects separators, then it searches 
upwards to find the first row with the same number of columns. If that 
row is all character then it deems that as the column name row.  So if 
you start autostart on 1, it's already at the top and it might catch the 
right separator by avoiding the data rows for separator detection.

On 24.12.2012 11:52, Hideyoshi Maeda wrote:
> Thanks for the quick response.
>
> I wasn't sure if I understood you correctly, but isn't the problem
> the way that autostart finds separators?
>
> and in my example, it had headers, so I think it would need to start
> from row 2 wouldn't it, i.e. the first row that has non-header 
> values?
>
> Thanks
>
> On 24 Dec 2012, at 11:44, Matthew Dowle <mdowle at mdowle.plus.com> 
> wrote:
>
>>
>> Hi,
>>
>> Ah yes, haven't hooked up the sep override yet, apologies, will fix.
>> Maybe setting autostart to the row number of the header row 
>> (probably 1)
>> might work.
>>
>> Thanks,
>> Matthew
>>
>>
>> On 24.12.2012 11:08, Hideyoshi Maeda wrote:
>>> oups…forgot to add the output from the verbose part…here it is...
>>>
>>> Detected eol as \r\n (CRLF) in that order, the Windows standard.
>>> Starting format detection on line 30 (the last non blank line in 
>>> the
>>> first 30)
>>> Detected sep as '/' and 3 columns
>>> Type codes: 003
>>> Found first row with 3 fields occuring on line 1 (either column 
>>> names
>>> or first row of data)
>>> The first data row has some non character fields. Treating as a 
>>> data
>>> row and using default column names.
>>> Count of eol after pos: 1143699
>>> Subtracted 1 for last eol and any trailing empty lines, leaving
>>> 1143698 data rows
>>>   0.153s ( 21%) Memory map (quicker if you rerun)
>>>   0.000s (  0%) Format detection
>>>   0.095s ( 13%) Count rows (wc -l)
>>>   0.001s (  0%) Allocation of 1143698x3 result (xMB) in RAM
>>>   0.480s ( 66%) Reading data
>>>   0.000s (  0%) Bumping column type midread and coercing data 
>>> already read
>>>   0.002s (  0%) Changing na.strings to NA
>>>   0.731s        Total
>>>
>>>
>>> On 24 Dec 2012, at 11:04, Hideyoshi Maeda 
>>> <hideyoshi.maeda at gmail.com> wrote:
>>>
>>>> Hi Matthew,
>>>>
>>>> I am using the new `data.table` `fread()` function to read my csv 
>>>> files, which has the format as follows when using the read.csv 
>>>> function
>>>>
>>>>           Date.and.Time Open High  Low Close Volume
>>>>   1 2007/01/01 22:51:00 5683 5683 5673  5673     64
>>>>   2 2007/01/01 22:52:00 5675 5676 5674  5674     17
>>>>   3 2007/01/01 22:53:00 5674 5674 5673  5674     42
>>>>
>>>> The value of the first column is all of: `2007/01/01 22:53:00`, 
>>>> the next 5 columns are separated with commas.
>>>>
>>>> but when reading the same file using fread i get the following 
>>>> output
>>>>
>>>>       V1 V2                                             V3
>>>>   1 2007  1 01 22:51:00,5683.00,5683.00,5673.00,5673.00,64
>>>>   2 2007  1 01 22:52:00,5675.00,5676.00,5674.00,5674.00,17
>>>>   3 2007  1 01 22:53:00,5674.00,5674.00,5673.00,5674.00,42
>>>>
>>>> This is because the autodetect is using the "/" as a separator...
>>>>
>>>> I tried overriding this using the `sep=","` argument but this does 
>>>> not seem to be used in the function anywhere.
>>>>
>>>> Furthremore when using verbose I get the following output, which 
>>>> suggests that I was right in thinking that "/" is used as a 
>>>> separator rather than ",".
>>>>
>>>> Is there any way to fix this, so that it correctly reads all 6 
>>>> columns separately?
>>>>
>>>> Thanks
>>>>
>>>> HLM
>>>>
>>>> On 21 Dec 2012, at 18:28, Matthew Dowle <mdowle at mdowle.plus.com> 
>>>> wrote:
>>>>
>>>>>
>>>>> Hi datatablers,
>>>>>
>>>>> Feedback and bug reports much appreciated :
>>>>>
>>>>> =====
>>>>> New function fread(), a fast and friendly file reader.
>>>>> * header, skip, nrows, sep and colClasses are all auto detected.
>>>>> * integers>2^31 are detected and read natively as 
>>>>> bit64::integer64.
>>>>> * accepts filenames, URLs and "A,B\n1,2\n3,4" directly
>>>>> * new implementation entirely in C
>>>>> * with a 50MB .csv, 1 million rows x 6 columns :
>>>>>  read.csv("test.csv")                                   # 30-60 
>>>>> sec
>>>>>  read.table("test.csv",<all known tricks, known nrows>) #    10 
>>>>> sec
>>>>>  fread("test.csv")                                      #     3 
>>>>> sec
>>>>> * airline data: 658MB csv (7 million rows x 29 columns)
>>>>>  read.table("2008.csv",<all known tricks, known nrows>) #   360 
>>>>> sec
>>>>>  fread("2008.csv")                                      #    50 
>>>>> sec
>>>>> See ?fread. Many thanks to Chris Neff and Garrett See for ideas,
>>>>> discussions and beta testing.
>>>>> =====
>>>>>
>>>>> 1.8.7 is passing checks on Unix and Windows (but not Mac yet) :
>>>>>
>>>>> install.packages("data.table", 
>>>>> repos="http://R-Forge.R-project.org")
>>>>> require(data.table)
>>>>> ?fread
>>>>> fread("your biggest baddest file")
>>>>>
>>>>> Oddly, R-Forge appears to be compiling Win64 with -O2 
>>>>> optimization rather
>>>>> than -O3 (but -O3 on Win32 ok), so speedups might not be as great 
>>>>> on Win64
>>>>> until that can be resolved on R-Forge, unless you compile 
>>>>> yourself. -O3
>>>>> has some optimizations that fread may benefit from. But 
>>>>> interested to hear.
>>>>>
>>>>> Seasons greatings!
>>>>>
>>>>> Matthew
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> datatable-help mailing list
>>>>> datatable-help at lists.r-forge.r-project.org
>>>>> 
>>>>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
>>>>
>>