These fast access methods are static. If the table or list is changed it may have to be re-indexed. Application programs should make sure that secondary indexes are updated to reflect changes made to their associated table or list files.

Indexing all of the columns separately in a table can be done by linking the table to as many different names as there are columns to index. Then each column can be indexed on one of the links.

-fc

c (not TAB) is the input column separation character.

-Dn

turn on debugging at level n, or 1 if n is not specified.

-ms

sequential method. index does nothing in this case because search simply scans through the whole file, first row to last.

-mr

record method. index builds a secondary file named tableorlist.r, which contains a fixed length column of row offsets.

-mb

binary method. index sorts tableorlist by the specified keycolumns. By default, keycolumns indexed with the binary method are sorted without regard to leading spaces (-b sort option). Other options applicable to both binary and inverted methods of indexing and searching modify the way in which the tableorlist is sorted, and are as follows:

-x

this option allows for searching whole or partial keys without regard to case, by applying the -f option to the sort command.

-n

this option sorts the keycolumns numerically instead of alphabetically, for subsequent numeric searches: for example, 001 is the same as 1 is the same as 1.0. When one of these options is used on the index command line, it must also be used on the search command line.

-mi

inverted method. This method is identical to binary except that index builds a secondary file of keys and offsets and sorts the keys in the secondary file rather than the table itself. The secondary file is named tableorlist.i. By default, index -mi strips leading, trailing and multiple blanks from the keys before adding them to the secondary table so that `Mr. Goober' is sorted in the same way as ` Mr. Goober '. The inverted method can get away with altering the keycolumns to provide a greater degree of conformity for search purposes because it's not actually touching the tableorlist itself, as in the binary method, which must rely solely on the final sort command to manipulate the keycolumns. See -mb above for an explanation of the other index options available with -mi.

-mh

hash method. index builds a secondary file, named tableorlist.h, and applies a simple algorithm to the keys in each row of tableorlist to produce a number which is within the bounds of the number of rows in the secondary file (called hashing). index writes the offset of the tableorlist row at the row number in the secondary file. If there's already an offset at that position, index probes each subsequent row in the index until it finds an empty slot at which to dump its offset. Because duplicate keys hash to the same row number, using the hash index and search method for tables which contain many identical or similar keys (lists of numbers, for example) can result in a slow sequential search rather than a fast hash search. The efficiency of hashing relies on the degree of uniqueness of keys and thus is very good for indexing data which varies widely. One way to determine how well-suited the hash method is for a particular table or key is to examine the resulting tablorlist.h file. An excellent result is one in which very few non-zero offsets appear in sequential rows; a not-so-good result is one in which big clumps of non-zero offsets appear in sequential rows.

Because the row number is based on the content of the key column, the way a key will be searched defines the way it should be indexed. This also means that the (-x, +x, -i) switches used on the index command line must also be used on the search command line, so that the key being searched can be processed, hashed and compared in the same way as the key it is seeking. By default, index -mh strips keys of leading, trailing and multiple internal blanks prior to hashing. This allows one to find keys which are word-for-word identical even though they may not be byte-for-byte matches. Also, blank key values are all assigned a null value in the hash table so that they can be found regardless of their (invisible) content. Options which alter the way hash tables are created are:

+x

this option prohibits changing the key as it appears in the table prior to hashing, forcing an exact match requirement of subsequent searches, extraneous blanks and all.

-x

this option hashes an all lower case version of the key in addition to trimming it as described above. When keys are searched also using -x, the result is case insensitivity: in other words, ` Mr. gOOber ' matches `Mr. Goober.' +x and -x are mutually exclusive options: whichever one appears first on the command line is the one which determines the indexing behavior.

-i

this option prohibits index from adding offsets of rows which contain blank keys to the index table. This is a good idea if there are many blank keys in the table because they will all hash to the same value, creating a long list of sequential offsets in the hash table. The side effect of -i is that search will not be able to find the blank keys. If this is a problem, a different method should be employed to index the table or to locate blank keys.

-n

this option hashes the string representation of keycolumns after applying the atof() function to them. Thus, keycolumns which contain any non-numeric notation will evaluate to zero, or be considered blank. Care should be taken when using this option, especially if the -i option is also being used. In general, the binary and inverted methods are better choices for numeric keys.

$ cat chart
Account Name
------- ----
100     Assets
101     Cash
111     Accounts Receivable
115     Notes Receivable
120     Deposits

Access to the data in chart can be sped up by indexing it and using the search command to find the row you want. To index chart for the inverted search method:

$ index -mi chart Account
$ cat chart.i
   Offset       Account
---------       -------
       26       100
       37       101
       46       111
       70       115
       91       120