Skip to content

An R package for reading Log Ascii Standard (LAS) files for well log data.

License

Notifications You must be signed in to change notification settings

donald-keighley/lasr

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

41 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

lasr

Donny Keighley 6/8/2022

lasr is a package designed for reading and writing Log Ascii Standard (LAS) files in R. Currently it is in the beta testing stages. As such, it is subject to significant ongoing changes and is not complete. For instance it can’t write LAS files yet…

Goals

lasr is primarily designed to import LAS files at high speed and in large batches. To accomplish this, most of it is written in C++ and connected to R with Rcpp. It stores the data in lists of data.table’s for fast manipulation.

Currently, the focus is on supporting reading LAS 3.0 as fully as possible. There will be some effort to handle non-standard LAS files but nothing too cute. lasr is being written to load files using as little information from the header as possible which should alleviate many common issues. Beyond that, the aim is to output helpful error messages so that non-standard files can be fixed. Afterall, one of the great things about LAS files is that they are human readable and can often be fixed using a simple text editor.

lasr is not designed to accomplish traditional petrophysical workflows and there are no plans to do so. There’s plenty of industry software for that. lasr is intended as a building block to facilitate people doing new and creative things with large volumes of log data. It’s also intended to help build up the library of geoscience packages for R.

For a more traditional approach, or if you’d prefer to use Python, you should check out the excellent lasio package.

Installation

You can install lasr from github using the install_github function from the devtools package. You will need to install RTools first since the package needs compilation.

if (!require('devtools')) install.packages('devtools')
library(devtools)
install_github('https://github.com/donald-keighley/lasr')

Currently, the only function is read.las which will import a vector of LAS file paths into a multi-part list. Each section of the file is stored as a separate element. In order to accomodate LAS 3.0 files which may have multiple log data sections, the log parameter, log definition, and log data are combined into numbered log elements. If your vector of paths contains more than one file, the output list will have an element for each file.

Here is an example reading a single LAS file that is included with the package:

library(lasr)
las = read.las(system.file("extdata", "las_3_cwls.las", package = "lasr"))

#Display the WELL section
head(las$well, 10)
##     mnemonic unit                    value           comment     format
##  1:     STRT    M                1660.1250 First Index Value           
##  2:     STOP    M                1660.8750  Last Index Value           
##  3:     STEP    M                   0.1250              STEP           
##  4:     NULL                       -999.25        NULL VALUE           
##  5:     COMP          ANY OIL COMPANY INC.           COMPANY           
##  6:     WELL         ANY ET AL 01-02-03-04              WELL           
##  7:      FLD                       WILDCAT             FIELD           
##  8:      LOC                    1-2-3-4W5M          LOCATION           
##  9:     SRVC      ANY LOGGING COMPANY INC.   SERVICE COMPANY           
## 10:     DATE                    13/12/1986      Service DATE DD/MM/YYYY
##     association
##  1:            
##  2:            
##  3:            
##  4:            
##  5:            
##  6:            
##  7:            
##  8:            
##  9:            
## 10:
#Display the log curves
head(las$log$log.1$data, 10)
##        DEPT     DT DPHI NPHI      YME                     CDES NMR[1] NMR[2]
## 1: 1660.125 123.45 0.11 0.37 1.45E+12          DOLOMITE W/VUGS     10     12
## 2: 1660.250 123.45 0.12 0.36 1.47E+12                LIMESTONE     12     15
## 3: 1660.375 123.45 0.13 0.35 2.85E+12            LOST INTERVAL     18     25
## 4: 1660.500 123.45 0.14 0.34 2.85E+12            LOST INTERVAL     18     25
## 5: 1660.625 123.45 0.15 0.33 2.85E+12            LOST INTERVAL     18     25
## 6: 1660.750 123.45 0.16 0.32 2.85E+12 SANDSTONE, SHALE STREAKS     18     25
## 7: 1660.875 123.45 0.17 0.31 2.85E+12            LOST INTERVAL     18     25
##    NMR[3] NMR[4] NMR[5]
## 1:     14     18     13
## 2:     21     35     25
## 3:     10      8     17
## 4:     10      8     17
## 5:     10      8     17
## 6:     10      8     17
## 7:     10      8     17

Most LAS files are version 2 and only have one log data section. If you know this is the case you can set flatten = TRUE and only the first log section will be returned. This makes referencing the log data quicker.

las = read.las(system.file("extdata", "las_3_cwls.las", package = "lasr"), flatten=TRUE)
head(las$log$data, 10)
##        DEPT     DT DPHI NPHI      YME                     CDES NMR[1] NMR[2]
## 1: 1660.125 123.45 0.11 0.37 1.45E+12          DOLOMITE W/VUGS     10     12
## 2: 1660.250 123.45 0.12 0.36 1.47E+12                LIMESTONE     12     15
## 3: 1660.375 123.45 0.13 0.35 2.85E+12            LOST INTERVAL     18     25
## 4: 1660.500 123.45 0.14 0.34 2.85E+12            LOST INTERVAL     18     25
## 5: 1660.625 123.45 0.15 0.33 2.85E+12            LOST INTERVAL     18     25
## 6: 1660.750 123.45 0.16 0.32 2.85E+12 SANDSTONE, SHALE STREAKS     18     25
## 7: 1660.875 123.45 0.17 0.31 2.85E+12            LOST INTERVAL     18     25
##    NMR[3] NMR[4] NMR[5]
## 1:     14     18     13
## 2:     21     35     25
## 3:     10      8     17
## 4:     10      8     17
## 5:     10      8     17
## 6:     10      8     17
## 7:     10      8     17

Speed Test

Since the purpose of this package is to load LAS files as quickly as possible, a speed test is included here with a comparison to python’s lasio. First, download a test dataset from the KGS website. In this case we’re using the 2016 logs data. Download and unzip them into a folder called “C:/temp/logs”, or modify the code for wherever you put it.

Next, import the first 500 files. We’ll use 4 threads for this comparison, although if you have more cores you can increase the number of threads to speed it up further. Only use this option if you are importing more than a handful of files, otherwise the parallel overhead will slow it down.

files = list.files('C:/temp/logs', pattern = '.las?', full.names=TRUE)
start.time = Sys.time()
las = read.las(files[1:500],nthreads=4)
end.time = Sys.time()
time.taken = end.time - start.time
time.taken
## Time difference of 25.96407 secs

Now in Python in parallel using 4 cores:

import lasio, glob, datetime, multiprocessing
from joblib import Parallel, delayed

num_cores = 4
files = glob.glob('C:/temp/logs/*.las')
start_time = datetime.datetime.now()
if __name__ == "__main__":
    las = Parallel(n_jobs=num_cores)(delayed(lasio.read)(file) for file in files[0:499])
end_time = datetime.datetime.now()
print('Duration: {}'.format(end_time - start_time))
## Duration: 0:04:45.961226

Clearly, lasr is faster, however, please don’t take this as a shot at lasio. The primary goal of this package is speed, and as such countless hours have been put into speed testing, de-bottlenecking, and enduring the pain of writing in C++. As with anything, there are tradeoffs, and lasr errs toward speed where lasio tends more toward user convenience. They are simply different products.

Good luck, and if you have any suggestions reach out!

About

An R package for reading Log Ascii Standard (LAS) files for well log data.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published