When a data set has many attributes and types include enumerated domains with many enumerations, preparing the Detailed_Description section of metadata under Entity_and_Attribute_Information is a laborious task. DBFmeta can be a lifesaver in this situation.

The statistical report produced by DBFmeta is valuable for detecting dirt in attributes, and to provide information that may be useful to improve or modify the metadata snippet produced by the tool. By default, the tool sends an abbreviated statistical report to the console, but a detailed statistical report may also be output as a user option. Either of those statistical reports may be redirected to an output file. Value occurence frequency analysis performed by the tool will identify unwanted duplication of values and subtle unwanted value differences caused by misspelling, bad entry, non-entry, and flags in numeric fields. It also reports the maximum and minimum non-flag values and the number of positive, negative, and zero values in numeric fields.

DBFmeta quickly creates a metadata snippet including each attribute and the Domain Values that it has guessed based on the data type of the attribute and the statistics of the attribute values. The tool is wise enough to include in its metadata snippet output the essential useful metadata elements for a complete Detailed_Description framework that only the metadata creator can know (the value for Attribute_Definition, Enumerated_Domain_Value_Definition, Attribute_Units_of_Measure) without also including the mindless mandatory elements (such as Enumerated_Domain_Value_Definition_Source and Attribute_Definition_Source) in the snippet that few metadata creators would include. By default, the tool also produces metadata elements and values describing the data type of the attribute according to the Arc8 ESRI profile, but as a user option (-strict), this output can be suppressed. The DBFmeta produced snippet may subsequently be pasted (using tkme for example) into a metadata document under the Entity_and_Attribute_Information section.

DBFmeta assigns attributes into only two types, Enumerated Domain and Range Domain. There are no statistical tests that the tool could perform to distinguish between Enumerated, Codeset, and Unrepresentable Domains. Defaulting to Enumerated Domain for character fields is an efficient mechanism because all the values that occur are reported in case they are needed for the metadata. The metadata creator can readily substitute the two compact domain representations for the longer Enumerated Domain when appropriate.

If the attribute is character type, it is assigned as enumerated domain, all unique values are given as alphabetically sorted enumerations in the metadata snippet, and the frequency of each enumeration is reported in the detailed statistics report. If the attribute is a Boolean field it is are reported as string type and enumerated as 'T', 'F' or '?' in the case of true, false, and empty respectively. If the attribute is numeric, the situation can be much more complex - the attribute may be guessed as and reported as range domain only, range domain with a flag enumeration, or enumerated domain. How well DBFmeta guesses is somewhat dependent on the number of records it has to test - the more it has, the better it will guess.

If the attribute type is integer and there isn't a lot of variety in the values, the attribute is assumed to be enumerated domain and the enumerations are listed in the metadata snippet. If there are only one or two negative values, those are assumed to be flag values (like -999), written as eumerations, and the remaining numbers greater than or equal to zero are written as range domain. All other cases of integers are written as pure range domain. Date type fields are analyzed as 8 digit integers.

If the atrribute type is floating point, it will be reported as a pure range domain field and the minimum and maximum values observed will be reported as the Range_Domain_Minimum and Range_Domain_Maximum (strictly speaking this is not in accordance with the definitions for those elements, but it is usually impossible to satisfy the CSDGM definition for those elements, and the actual observed minimum and maximum values have a lot of value).

The interactive mode of operation is not recommended, especially for datafiles with many attributes. The user is only prompted for attribute definitions and for units of range domain values. This (thankfully) means the user must use an alternate method to fill in definitions for enumerations. It is impossible to restart an interactive session with DBFmeta once interrupted (by a ^C, say, or a power glitch), one must start over from the beginning with the first attribute.

To make the tool readily usable from a DOS command window, the administrator need only insure the location of the dbfmeta.exe is indicated by the PATH environment variable or that the tool is launched via a batch file whose location is identified in the PATH. Most administrators would benefit from periodically downloading and installing the complete metadata tool suite to insure the most up to date version of all of the metadata tools developed by Peter Schweitzer.

• When supplied with an input filename for a file that does not have a .dbf extension (even though its content is that of a valid .dbf file), DBFmeta returns is 'Error: could not open input file file_name_without_dbf_ext.'
• When presented with an input file with a .dbf extension that is not actually a valid .dbf file, DBFmeta sends a long stream of WARNING messages to console which (in the case tested) eventually terminated in an error message from DBFmeta announcing it was having trouble allocating space.