Data Storage

«  Referencing and Fitting   ::   Contents   ::   Calibration and Concentration Determination  »

Data Storage

Processed data files can be stored on-disk using the GcmsStore object located in the gcmstools.datastore module. Not only does this create a convenient storage solution for processed data sets, it is also necessary when running calibrations on a group of related data sets. The file is a HDF file, which is an open-source high performance data storage container optimized for numerical data. Creation and manipulation of this file is controlled using a combination of two Python libraries: PyTables and Pandas. PyTables provides a high-level interface to create and modify HDF files, and Pandas is a very powerful package for working with tabular data. Both of these project have extensive documentation of their many advanced features, so little detail on their usage is provided here.

About GcmsStore Implementation

The GcmsStore object is a subclass of the Pandas HDFStore object, so it will contain all of the functions described for that object. The GcmsStore class simply adds a number of custom functions specific for GCMS data sets.

Create/Open the Container

A gcmstools GcmsStore object must be created with a file name argument. If a file with this name already exists, it will be opened for appending or modification. The default behavior is to compress all the data going into this file using the ‘blosc’ compression library and the highest compression level (9). See the Pandas HDFStore documentation for other accepted keyword arguments, especially the compression arguments if different values are required.

In : from gcmstools.datastore import GcmsStore

In : h5 = GcmsStore('data.h5')

Closing the File

In general, you will want to close the HDF file when you’re done, although this is not strictly necessary.

In : h5.close() # Only do this when you're done

Recompressing the HDF File

HDF files are designed to be written once and read many times. If you are repeatedly adding new files to the HDF storage container, the file size may become much larger than seems necessary. You can recompress the file using the compress method (which first closes the HDF file).

In : h5.compress() # This closes the file as well.

Adding Data

Added files to this storage container is done using the append_gcms method, which can take either a single data object or a list of objects, if you have many objects to add at one time.

In : h5.append_gcms(data)
HDF Appending: datasample1.CDF

In : h5.append_gcms([otherdata1, otherdata2])
HDF Appending: otherdata1.CDF
HDF Appending: otherdata2.CDF

Data files can be added at any stage of the processing chain; however, the calibration process will not work properly if you don’t reference/fit the data first. You can add an already existing data file as well. The GcmsStore object will check if that file is different than the saved version before overwriting the existing object. If it is not changed, then the file will be skipped.

Viewing the File List

You can see a list of the files that are stored in this file by viewing the files attribute, which is a Pandas DataFrame.

In : h5.files
          name         filename
0  datasample1  datasample1.CDF
1  otherdata1   otherdata1.CDF
2  otherdata2   otherdata2.CDF

There are two name columns in this table: “name” and “filename”. The latter is the full file and path name as given when the GcmsFile object was created. Keep in mind that the path information may not be correct if you’ve moved the location of this storage file. In order to efficiently store the data on disk, the full file name is internally simplified the “name”. This simplification removes the path and file extension from the file name. In addition, it replaces all ”.”, “-”, and spaces characters with “_”. If the file name starts with a number, the prefix “num” is added.


You will encounter problems if two or more file names simplify to the same “name”. However, if you’re file naming system does not produce unique file names for different data sets, you will most certainly have more problems in the long run than just using these programs.

Extracting Stored Data

You can extract data from the storage file using the extract_gcms method. This function takes one argument which is the name of the dataset that you want to extract. This name can be either the simplified name or the full filename (with or without the path). The extracted data is the same file object type as you stored originally.

In : extracted = h5.extract_gcms('datasample1')

In : extracted.filetype
Out: "AiaFile"

Stored Data Tables

This HDF data file may contain a number of Pandas data tables (DataFrames) with information about the files, calibration, etc. A list of currently available tables can be obtained by directly examining the GcmsStore by directly examining the GcmsStore instance. (Note: you won’t see these attributes using tab completion.)

In : h5
<class 'pandas.io.pytables.HDFStore'>
File path: data.h5
/calibration            frame        (shape->[6,8])
/calinput               frame        (shape->[30,9])
/datacal                frame        (shape->[49,6])
/files                  frame        (shape->[1,2])

Directly viewing these tables is trivial.

In : h5.calibration
               Start  Stop  Standard         slope      intercept         r  \
benzene          2.9   3.5       NaN  38629.931565 -367129.586850  0.998767
phenol          14.6  15.1       NaN  30248.192619   65329.897933  0.999136

                      p       stderr
benzene        0.000052  1108.344872
phenol         0.000030   726.257380

More information on using these tables is provided in Appendix B.

«  Referencing and Fitting   ::   Contents   ::   Calibration and Concentration Determination  »