Welcome to dliswriter - an open-source Python package for writing DLIS files.

The package allows you to specify the structure, data, and metadata of your DLIS file in a simple and flexible fashion. A minimal example is shown below.

import numpy as np  # for creating mock datasets
from dliswriter import DLISFile, enums

df = DLISFile()

df.add_origin("MY-ORIGIN")  # required; can contain metadata about the well, scan procedure, etc.

# define channels with numerical data and additional information
n_rows = 100  # all datasets must have the same number of rows
ch1 = df.add_channel('DEPTH', data=np.linspace(0, 10, n_rows), units=enums.Unit.METER)
ch2 = df.add_channel("RPM", data=np.arange(n_rows) % 10)
ch3 = df.add_channel("AMPLITUDE", data=np.random.rand(n_rows, 5))

# define frame, referencing the above defined channels
main_frame = df.add_frame("MAIN-FRAME", channels=(ch1, ch2, ch3), index_type=enums.FrameIndexType.BOREHOLE_DEPTH)

# write the data and metadata to a physical DLIS file
df.write('./new_dlis_file.DLIS')

For more details about the DLIS file format and using dliswriter, please see the documentation.

According to our rough measurements, the file writing time seems to be pretty much linearly dependent on the amount of data, in particular the number of rows. There is also some dependency on the dimensionality of the data - e.g a single image (2D dataset) with 1000 columns will write about 20% faster than 10 images of 100 columns each. A rough estimate of the writing speed is about 20M float64 values per second (measured on an x64-based PC with Intel Core i9-8950HK with MS Windows 11 and Python 3.9.19).

The performance may be further tuned by adjusting the input_chunk_size and output_chunk_size of the writer (see this example). The former limits how much of the input data are loaded to memory at a time; the latter denotes the number of output bytes kept in memory before each partial file write action. The optimal values depend on the hardware/software configuration and the characteristics of the data (number and dimensionality of the datasets), but the defaults should in general be a good starting point.

Please note that some DLIS viewer applications are not fully compliant with the DLIS standard. If a DLIS file produced by dliswriter causes issues in some of the viewers, it might not necessarily be a dliswriter bug. Some of the known compatibility issues - and ways of dealing with them - are described in a dedicated section of the documentation. If you run into problems not covered by the documentation, please open a new issue.

dliswriter can be installed from PyPI:

pip install dliswriter

or Anaconda (in a conda environment):

conda install dliswriter -c conda-forge

Setting up dliswriter for development purposes requires:

• Python (at least 3.9)
• Anaconda, e.g. Miniconda
• git

Once these requirements are fulfilled, follow the steps below:

1. Clone the repository and enter it. From a console:

   git clone https://github.com/well-id/dliswriter.git
   cd dliswriter
   

2. Create the dlis-writer environment from the environment.yaml file and activate it:

   conda env create -f environment.yaml
   conda activate dlis-writer
   

3. Install dliswriter in editable mode using pip:

   pip install --no-build-isolation --no-deps -e .
   

   For explanation of the required flags, see this issue.

4. You're good to go! For verification, you can run the tests for the package (still from inside the dliswriter directory):

   pytest .
   

To contribute to the dliswriter, please follow this procedure:

1. Fork the repository
2. Clone the fork to your computer
3. Check out the devel branch: git checkout devel
4. Create a new branch from devel: git checkout -b <your branch name>
5. Make your changes, commit them, and push them
6. From the GitHub page of your fork, create a pull request to the original repository.

You can find some more detailed instructions about the fork-and-pull request workflow in the GitHub Docs.

You might also want to have a look at our issues log.

dliswriter has been developed at Well ID by:

• Dominika Dlugosz
• Magne Lauritzen
• Kamil Grunwald
• Omer Faruk Sari

Based on the definition of the RP66 v1 standard.