Method and apparatus for indexing database columns with bit vectors

An index method of data management in a data processing system, the method comprising:

• providing an application program, a database management system and a server;
• forming an index of bit vectors in the server, the index being formed from an ordered set of data values disposed in a database of the database management system as a plurality of data records arranged in row and column format, each data value disposed in the database having a bit pattern of arbitrary length and an identifier for a particular column associated therewith, the index being formed by:
• (i) creating a number of bit vectors, wherein the number of bit vectors is equal to the length of the longest of the bit patterns for the data values and the length of each bit vector is equal to the number of data values;
• (ii) assigning to a unique position in successive bit vectors each bit of the bit pattern for each data value, wherein each bit of the bit pattern for each data value is assigned to identical unique position in each of the successive bit vectors;
• so that each bit vector comprises an array storing a transposed sequence of bit values derived by taking a single bit value for each record of said database table, each bit vector comprising bit values taken from a single bit position for said particular column of said database table;
• receiving a search value and a search function from the application program;
• comparing the search value to the data values assigned into the bit vectors by manipulation of the bit vectors based on the search value and the search function;
• producing a resultant bit vector having a bit pattern indicating the identifier of one or more data values, thereby indicating that said one or more data values satisfy the search function;
• and
• compressing the ordered set of data values disposed in the database by deleting all data values which can be derived from said index of bit vectors.

A database management system comprising:

• a server having a database comprising an ordered set of data values stored as a plurality of data records arranged in row and column format, each data value having a bit pattern and an identifier associated therewith, the server comprising a memory, and a processor;
• means for creating a number of bit vectors in the memory, such that the number of bit vectors is equal to the length of the longest of the bit patterns for the data values and the length of each bit vector is equal to the number of data values;
• means for accessing one of the ordered set of data values stored in the database;
• means for assigning each successive bit of the bit pattern for the accessed data value to a unique position in successive bit vectors, the bits being assigned into an identical unique position in each of the successive bit vectors, so that each bit vector comprises an array storing a column of bit values created by taking a single bit value at a particular unique bit position of one column for each record of said database table;
• means for repeating the accessing and assigning steps for each remaining data value to form an index of bit vectors for the data values;
• and
• means for compressing the ordered set of data values disposed in the database by eliminating data values which can be derived from said bit vectors.

A method for data management in a data processing system comprising the steps of:

• storing a database comprising data arranged as an ordered set of data values in row and column format, each data value having a bit pattern of arbitrary length and an identifier associated therewith;
• creating a number of bit vectors, wherein the number of bit vectors is equal to the length of the longest of the bit patterns for the data values and the length of each bit vector is equal to the number of data values, said creating step comprising substeps of:
• (i) accessing one of the ordered set of data values stormed in the database;
• (ii) assigning to a unique position in successive bit vectors each bit of the bit pattern for said data value, each bit being assigned to an identical unique position in each of the successive bit vectors;
• and
• (iii) repeating the accessing and assigning steps for each remaining data value to form an index of bit vectors for the data values, so that each bit vector comprises an array storing a column of bit values created by taking a single bit value at a particular unique bit position of one column for each row of said database;
• receiving a search value and a search function from the application program;
• comparing the search value to the data values assigned into the bit vectors by manipulation of the bit vectors based on the search value and the search function;
• producing a resultant bit vector having a bit pattern indicating the identifier of one or more data values, thereby indicating that said one or more data values satisfy the search function;
• and
• reconstructing each data value associated with each identifier indicated in the bit pattern of the resultant bit vector by selecting each bit, in each successive bit vector of the index, having a unique position corresponding to said identifier.

A database management system comprising:

• a server having a database storing an ordered set of data values as rows of data records, each data value having a bit pattern and an identifier for a particular database column associated therewith, the server comprising a memory and a processor;
• means for creating a number of bit vectors in the memory, such that the number of bit vectors is equal to the length of the longest of the bit patterns for the data values and the length of each bit vector is equal to the number of data values, said means for creating including:
• means for accessing one of the ordered set of data values stored in the database;
• means for assigning each bit of the bit pattern for the accessed data value to a unique position in successive bit vectors, the bits being assigned into an identical unique position in each of the successive bit vectors;
• and
• means for repeating the accessing and assigning steps for each remaining data value to form an index of bit vectors for the data values, so that each bit vector comprises an array storing a column of bit values created by taking a single bit value at a particular unique bit position of one column for each row of said database;
• means for receiving a search value and a search function from the application program;
• means for comparing the search value to the data values assigned into the bit vectors by manipulation of the bit vectors based on the search value and the search function;
• means for producing a resultant bit vector having a bit pattern indicating the identifier of one or more data values, thereby indicating that said one or more data values satisfy the search function;
• and
• means for reconstructing each data value associated with each identifier indicated in the bit pattern of the resultant bit vector by selecting each bit, in each successive bit vector of the index, having a unique position corresponding to said identifier.

In a database system comprising a database storing information in a database table as a sequence of data records arranged in row and column format, an improved method for indexing said database table, the method comprising:

• selecting a particular column of the database table to index, said particular column representing a particular category of information for which each data record stores a data value, each data value representing information in the database stored as a sequence of bit values, each bit value having a particular bit position in said sequence of bit values;
• and
• creating an index for said particular column comprising a plurality of bit vectors by performing substeps of:
• (i) determining a maximum number of bit positions employed in said sequence of bit values for storing information for said particular column,
• (ii) initializing a sequence of bit vectors based on said determined maximum number of bit positions, so that each bit vector is associated with a particular one of said bit positions employed, and
• (iii) copying data values stored for said particular column by said sequence of data records into said sequence of bit vectors, so that for each particular bit position at each particular data record any bit value stored thereat is copied into the bit vector associated with that bit position, the bit value being copied into a particular bit position of the bit vector which corresponds to position of the particular data record relative to other data records in said sequence of data records.