Search-on-the-fly with merge function

[0001] This non-provisional application claims the benefit of U.S. provisional patent Application No. 60/227,305, entitled “SEARCH-ON-THE-FLY WITH MERGE FUNCTION,” filed on Aug. 24, 2000. The provisional application is hereby incorporated by reference.

[0002] This application is a continuation-in-part of application Ser. No. 09/513,340, filed Feb. 25, 2000, entitled Search-On-The-Fly/Sort-On-The-Fly Search Engine, which is hereby incorporated by reference.

[0003] The technical field is information management systems, interfaces, and mechanisms, and methods for searching one or more databases.

[0004] In the most general sense, a database is a collection of data. Various architectures have been devised to organize data in a computerized database. Typically, a computerized database includes data stored in mass storage devices, such as tape drives, magnetic hard disk drives and optical drives. Three main database architectures are termed hierarchical, network and relational. A hierarchical database assigns different data types to different levels of the hierarchy. Links between data items on one level and data items on a different level are simple and direct. However, a single data item can appear multiple times in a hierarchical database and this creates data redundancy. To eliminate data redundancy, a network database stores data in nodes having direct access to any other node in the database. There is no need to duplicate data since all nodes are universally accessible. In a relational database, the basic unit of data is a relation. A relation corresponds to a table having rows, with each row called a tuple, and columns, with each column called an attribute. From a practical standpoint, rows represent records of related data and columns identify individual data elements. The order in which the rows and columns appear in a table has no significance. In a relational database, one can add a new column to a table without having to modify older applications that access other columns in the table. Relational databases thus provide flexibility to accommodate changing needs.

[0005] All databases require a consistent structure, termed a schema, to organize and manage the information. In a relational database, the schema is a collection of tables. Similarly, for each table, there is generally one schema to which it belongs. Once the schema is designed, a tool, known as a database management system (DBMS), is used to build the database and to operate on data within the database. The DBMS stores, retrieves and modifies data associated with the database. Lastly, to the extent possible, the DBMS protects data from corruption and unauthorized access.

[0006] A human user controls the DBMS by providing a sequence of commands selected from a data sublanguage. The syntax of data sublanguages varies widely. The American National Standards Institute (ANSI) and the International Organization for Standardization (ISO) have adopted Structured English Query Language (SQL) as a standard data sublanguage for relational databases. SQL comprises a data definition language (DDL), a data manipulation language (DML), and a data control language (DCL). The DDL allows users to define a database, to modify its structure and to destroy it. The DML provides the tools to enter, modify and extract data from the database. The DCL provides tools to protect data from corruption and unauthorized access. Although SQL is standardized, most implementations of the ANSI standard have subtle differences. Nonetheless, the standardization of SQL has greatly increased the utility of relational databases for many applications.

[0007] Although access to relational databases is facilitated by standard data sublanguages, users still must have detailed knowledge of the schema to obtain needed information from a database since one can design many different schemas to represent the storage of a given collection of information. For example, in an electronic commerce system, product information, such as product SKU, product name, product description, price, and tax code, may be stored in a single table within a relational database. In another electronic commerce system, product SKU, product name, description, and tax code may be stored in one table while product SKU and product price are stored in a separate table. In this situation, a SQL query designed to retrieve a product price from a database of the first electronic commerce system is not useful for retrieving the price for the same product in the other electronic system's database because the differences in schemas require the use of different SQL queries to retrieve product price. As a consequence, developers of retail applications accessing product information from relational databases may have to adapt their SQL queries to each individual schema. This, in turn, prevents their applications from being used in environments where there are a wide variety of databases having different schemas, such as the World Wide Web.

[0008] A further problem with conventional searches, search engines, data access and data retrieval is a tendency to return very large amounts of data, or to require the search parameters to be narrowed. When large amounts of data are presented, the display may take many “pages” before all data is seen by the user. The time and expense involved in such a data review may be significant, inconvenient, not user friendly or efficient.

[0009] Sort-on-the-Fly/Search-on-the-Fly data retrieval methods and apparatus (hereafter, search-on-the-fly) provide an intuitive means for accessing or searching databases, allowing a user to access or obtain information about data in the database without having to know anything about the database structure. Sort-on-the-Fly/Search-on-the-Fly is an information gathering process or analysis process about data stored in one or more databases. The on-the-fly methods and apparatus often use or include sorting and searching. While Sort-on-the-Fly/Search-on-the-Fly may be a search engine or part of a search engine, it may also stand alone or make calls to a search engine. For example, database search engines may be used in conjunction with on-the-fly methods and apparatus.

[0010] Using Sort-on-the-Fly/Search-on-the-Fly, a user selects a desired term, and the user is delivered all instances of the desired term, even if a specific file or table does not contain the instance. For example, if a user wants to enter a database using the name of a specific individual as a database entry point, a database manager or other software will access the database using the desired name, and will organize the results so that all entries associated with that name are displayed. The database need not have a specific file (in a flat database) or a table (in a relational database) of names. The user may perform further on-the-fly searches or information retrieval to narrow or focus the results, or for other reasons. For example, given results for all names that include the name “Smith,” the user may then decide to obtain information for all “Smiths” that include an association to an address in New Jersey. Search-on-the-fly then conducts a further information gathering using this criteria and produces a second result. Further narrowing or broadening of the analysis is permitted, with search-on-the-fly returning results based on any new criteria.

[0011] In an embodiment, search-on-the-fly uses graphical user interfaces (GUIs) and one or more icons to make the information gathering process as efficient as possible. The GUIs may incorporate one or more pull down menus of available sorting terms. As a user selects an item from a first pulldown menu, a subsequent pulldown menu displays choices that are available for sorting or searching. The process may be continued or repeated until Sort-on-the-Fly/Search-on-the-Fly has retrieved or displayed a discrete data entry from the database. The pulldown menus are not pre-formatted. Instead, the pulldown menus are created “on-the-fly” as the user steps through the sort and/or search process. Thus, search-on-the-fly is inherently intuitive, and allows a user with little or no knowledge of the database contents, its organization, or a search engine search routine to execute comprehensive analysis, sorting and/or searches that return generally accurate results.

[0012] Search-on-the-fly also searches on key words specified by the user. Search-on-the-fly can be used to exclude certain items. Search-on-the-fly incorporates other advanced features such as saving results by attaching a cookie to a user's computer, and associating icons with the results.

[0013] Search-on-the-fly may be used with both internal and external databases. For example, Search-on-the-fly may be used with a company internal database and one or more databases accessible through the Internet.

[0014] Search-on-the-fly is user-friendly. With one interface, many different types of databases or database schemas may be searched or sorted.

[0015] Finally, the search-on-the-fly technique, and other techniques discussed above may be used in conjunction with a method of doing business, particularly a business method that uses the Internet as a communications backbone.

[0036] Ordinary search engines place constraints on any search. In particular, a partial ordering of available search criteria limits application of the search engine only to certain search sequences. The user is given a choice of search sequences, and the order in which individual search steps in the search sequence become available limits the direction of the search. A user who desires to take a vacation cruise may use an Internet search engine to find a desired vacation package. The search begins with presentation of a list of general categories, and the user clicks on “travel,” which produces a list of subcategories. The user then clicks on “cruises” from the resulting list of subcategories, and so on in a cumulative narrowing of possibilities until the user finds the desired destination, date, cruise line, and price. The order in which choices become available amounts to a predefined “search tree,” and the unspoken assumption of the search engine designer is that the needs and thought processes of any user will naturally conform to this predefined search tree.

[0037] To an extent, predefined constraints are helpful in that predefined constraints allow a search engine to logically and impersonally order the user's thoughts in such a way that if the user has a clear idea of what object the user wants, and if the object is there to be found, then the user is assured of finding the object. Indeed, the user may want to know that choosing any available category in a search sequence will produce an exhaustive and disjunctive list of subcategories from which another choice can be made. Unfortunately, an unnecessarily high cost is too often paid for this knowledge: The user is unnecessarily locked into a limited set of choice sequences, and without sufficient prior knowledge of the object being sought, this limitation can become a hindrance. Specifically, where prescribed search constraints are incompatible with the associative relationships in the user's mind, a conflict can arise between the thought processes of the user and the function of the search engine.

[0038] At one time, such conflicts were written off to the unavoidable differences between computers and the human mind. However, some “differences” are neither unavoidable nor problematic. In the case of search engine design, the solution is elegant: upon selecting a category or entering a keyword, the user can be given not only a list of subcategories, but the option to apply previously available categories as well. In slightly more technical terms, the open topology of the search tree can be arbitrarily closed by permitting search sequences to loop and converge. Previous lists can be accessed and used as points of divergence from which new sub-sequences branch off, and the attributes corresponding to distinct sub-sequences can later be merged.

[0039] Sort-on-the-fly/search-on-the-fly data analysis, sorting access and retrieval methods and apparatus (hereafter, search-on-the-fly search engine) provide an intuitive means for analyzing various types of databases, allowing a user to obtain information about and/or access data in the database without having to know anything about the database structure. A user selects a desired term, and a database manager reviews the database for all instances of the desired term, even if a specific file or table does not contain the instance. For example, if a user wants to analyze the database using the name of a specific individual as a database entry point, the database manager will search the database or index using the desired name, and will organize the results so that all entries associated with that name are displayed. The database need not have a specific file (in a flat database) or a table (in a relational database) of names. The user may perform further on-the-fly searches to narrow the search results, or for other reasons. The search engine then conducts a further search using this criteria and produces a second search result. Further narrowing or broadening of the search are permitted, with the search engine returning results based on any new criteria.

[0040] This on-the-fly method or process can be used to simply analyze data or gather information about data stored in a database. The actual data itself does not need to be fetched, displayed, printed or even sorted. The user may simply wish to use this tool to “clean-up” data or understand how data could be sorted or for other reasons.

[0041] FIG. 1 is a block diagram of a system 10 that uses search-on-the-fly. In FIG. 1, a database 12 is accessed using a hardware/software interface device 100 to provide data to a user terminal 14. Additional databases 13 and 15 may also be accessed by the terminal 14 using the device 100. The databases 12, 13 and 15 may use different schemas, or may use a same schema. As will be described later, the device 100 may include the search-on-the-fly search apparatus. In an alternative embodiment, the search-on-the-fly search engine may be co-located with the terminal 14. In yet another embodiment, the search-on-the-fly search engine may be incorporated into the structure of one or more of the databases 12, 13 and 15. The device 100 may interface with any one or more of the databases 12, 13 and 15 using a network connection such as through the Internet, for example. Other communications mediums may also be used between the terminal 14, the device 100 and any one or more of the databases 12, 13 and 15. These mediums may include the public switched telephone network (PSTN), cable television delivery networks, Integrated Services Digital Networks (ISDN), digital subscriber lines (DSL), wireless means, including microwave and radio communications networks, satellite distribution networks, and any other medium capable of carrying digital data.

[0042] The system shown in FIG. 1 is but one of many possible variations. The search-on-the-fly search engine could also be incorporated within a single computer, such as a personal computer, a computer network with a host server and one or more user stations, an intranet, and an Internet-based system, as shown in FIG. 2. Referring again to FIG. 2, the terminal 14 may be any device capable of displaying digital data including handheld devices, cellular phones, geosynchronous positioning satellite (GPS) devices, wrist-worn devices, interactive phone devices, household appliances, televisions, television set top boxes, handheld computers, and other computers.

[0043] FIG. 3 is a detailed block diagram of an exemplary search-on-the-fly search engine 125. The search engine 125 includes a request analyzer 130 that receives search requests 114 from the terminal 14 (not shown in FIG. 3) and sends out updated requests 115 to a query generator 150. A status control 140 receives a status update signal 116 and a request status control signal 118 and sends out a request status response 119 to the request analyzer 130. The status control 140 also keeps track of search cycles, that is, the number of search iterations performed. The query generator 150 receives the updated requests 115 from the request analyzer 130 and sends a database access signal 151 to a database driver 170. The query generator 150 receives results 153 of a search of the database 12 (not shown in FIG. 3) from the database driver 170. The query generator 150 provides a display signal 175 to the terminal 14. The database driver 170 sends a database access signal 171 to the database 12. Finally, a database qualifier 160 receives information 161 from the database driver 170 and provides a list 163 of available data fields from the database 12. As will be described later, the list of available data fields 163 may be displayed to a user at the terminal 14, and may be sorted and processed using the request analyzer 130 in conjunction with the database qualifier 160. The database qualifier 160 also receives search information and other commands 131 from the request analyzer 130.

[0044] The search engine 125 may identify a database schema by simply using a trial and error process. Alternatively, the search engine 125 may use other techniques know in the art. Such techniques are described, for example, in U.S. Pat. No. 5,522,066, “Interface for Accessing Multiple Records Stored in Different File System Formats,” and U.S. Pat. No. 5,974,407, “Method and Apparatus for Implementing a Hierarchical Database Management System (HDBMS) Using a Relational Database Management System (RDBMS) ad the Implementing Apparatus,” the disclosures of which is hereby incorporated by reference.

[0045] The search engine 125 provides search-on-the-fly search capabilities and more conventional search capabilities. In either case, the search engine 125 may perform a preliminary database access function to determine if the user has access to the database 12. The search engine 125 also determines the database schema to decide if the schema is compatible with the user's data processing system. If the database schema is not compatible with the user's processing system, the search engine 125 may attempt to perform necessary translations so that the user at the terminal 14 may access and view data in the database 12. Alternatively, the search engine 125 may provide a prompt for the user indicating incompatibility between the terminal 14 and a selected database.

[0046] The search engine 125 may conduct a search using one or more search cycles. A search cycle includes receipt of a request 114, any necessary formatting of the request 114, and any necessary truncation steps. The search cycle ends when a result list 175 is provided to the terminal 14. The search engine 125 may retain a status of each past and current search cycle so that the user can modify the search at a later time. The user may also use this feature of retaining a status of past and current search cycles to combine results of multiple searches, using, for example, a Boolean AND function, a Boolean OR function, or other logic function. The above listed functions will be described in more detail later.

[0047] The search-on-the-fly function of the search engine 125 begins by determining available data fields of the database 12. The database 12 may have its data organized in one or more data fields, tables, or other structures, and each such data field may be identified by a data field descriptor. In many cases, the data field descriptor includes enough text for the user at the terminal 14 to determine the general contents of the data field. The list of data fields may then be presented at the terminal 14, for example, in a pull down list. An example of such a data field result list is shown in FIG. 4, which is from a federal database showing data related to managed health care organizations. This database is available at http://tobaccopapers.org/dnld.htm. In FIG. 4, the first data field listed is “PlanType,” which is shown in result list 156. Other data field descriptors show the general categories of data in the database.

[0048] Using the terminal 14, the user may select one of the data field descriptors to be searched. For example, the user could select “city.” If a number of entries, or records, in the city data field is short, a further result list of complete city names may be displayed. If the entries are too numerous to be displayed within a standard screen size, for example, the search engine 125 may, in an iterative fashion, attempt to reduce, or truncate, the result list until the result list may be displayed. In the example shown in FIG. 4, entries in the city data field are so numerous (the database includes all U.S. cities that have a managed health care organization) that the search engine 125 has produced a result list 157 that shows only a first letter of the city. Based on the available database data fields, the user may then perform a further search-on-the-fly. In this case, the user may choose cities whose first initial is “N.” The search engine 125 then returns a result list 158 of cities whose names start with the letter “N.” Because in this instance the result list 158 is short, no further truncation is necessary to produce a manageable list.

[0049] FIG. 5 is a more detailed block diagram of the request analyzer 130. A protocol analyzer 133 receives the request 114 and provides an output 135 to a constraint collator 136. The protocol analyzer 133 examines the received request 114, determines a format of the request 114, and performs any necessary translations to make the request format compatible with the database to be accessed. If the database to be accessed by the terminal 14 is part of a same computer system as the terminal 14, then the protocol analyzer 133 may not be required to perform any translations or to reformat the request 114. If the database to be accessed is not part of the same computer system as the terminal 14, then the protocol analyzer 133 may be required to reformat the request 114. The reformatting may be needed, for example, when a request 114 is transmitted over a network, such as the Internet, to a database coupled to the network.

[0050] The constraint collator 136 provides the updated request 115 (which may be an initial request, or a subsequent request) to the query generator 150. The constraint collator 136 is responsible for interpreting the request 114. The constraint collator 136 performs this function by comparing the request 114 against information stored in the status control 140. In particular, the constraint collator 136 sends the request status control signal 118 to the status control 140 and receives the request status response 119. The constraint collator 136 then compares the request status response 119 to constraint information provided with the request 114 to determine if the constraint status should be updated (e.g., because the request 114 includes a new constraint). In an embodiment, the constraint collator 136 compares constraint information in a current request 114 to constraint information residing in the status control 140, and if the current request 114 includes a new constraint, such as a new narrowing request (for example, when the user clicks, touches or points over a field shown in a last search cycle), then the constraint collator 136 adds the updated information and sends the updated request 115 to the query generator 150. If the constraint status should be updated, the constraint collator 136 sends the status update 118 to the status control 140. If the request 114 is a refresh request, the constraint collator 136 sends a reset command 131 to the database qualifier 160. The updated request 115 (possibly with a new constraint) is then sent to the query analyzer 150 for further processing.

[0051] FIG. 6 is a block diagram of the query generator 150. The overall functions of the query generator 150 are to scan a database, such as the database 12, using the database driver 170, and to collect search results based on constraints supplied by the request analyzer 130. The query generator 150 then returns the search results 175 to the terminal 14.

[0052] The query generator 150 includes a truncator 152 and a dispatcher 154. The truncator 152 receives the updated request 115, including a new constraint, if applicable. The truncator 152 creates new queries, based on new constraints, and applies the new requests 151 to the database 12 using the database driver 170. Many different methods of truncating for display or viewing may be used by truncator 152. The truncator 152 may include a variable limit 155 that is set, for example, according to a capacity of the terminal 14 to display the search results 175. If data retrieved from the database 12 exceed the limit value, the truncator 152 adjusts a size (e.g., a number of entries or records) of the data until a displayable result list is achieved. One method of adjusting the size is by cycling (looping). Other methods may also be used to adjust the size of the result list. For example, the terminal 14 may be limited to displaying 20 lines of data (entries, records) from the database 12. The truncator 152 will cycle until the displayed result list is at most 20 lines. In an embodiment, the truncation process used by the truncator 152 assumes that if the user requests all values in a particular data field from the database 12, and there are no other constraints provided with the request 114, and if the size of the resulting result list is larger than some numeric parameter related to a display size of the terminal 14, then the constraints may be modified by the truncator 152 so that the result list can accommodated (e.g., displayed on one page) by the terminal 14. For example, instead of a full name of a city, some part of the name—the first n letters—is checked against the database 12 again, and n is reduced until the result list is small enough for the capacity of the terminal 14. If the maximum number of displayable results is three (3), and the database 12 contains the names of six cities “Armandia, Armonk, New Orleans, New York, Riverhead, Riverdale,” then the first attempt to “resolve” the result list will stop after a result list display is created with the full name of the cities:

• [0053] Armandia, Armonk, New Orleans . . . (the limit was reached)
• [0054] Try again with 7 characters:
• [0055] Armandia, Armonk, New Orl, New Yor, (limit reached again)
• [0056] Again with 5 characters:
• [0057] Armandia, Armonk, New O, New Y, (limit reached again)
• [0058] Again with 3 characters:
• [0059] Arm ( . . . ), New ( . . . ), Riv ( . . . ). These results may now be displayed on the terminal 14.
  The display of Arm, New, Riv can then be used to conduct a further search-on-the-fly.
  For example, a user could then select Riv for a further search-on-the-fly. The result list returned would then list two cities, namely Riverhead and Riverdale.

[0060] In another embodiment, a fixed format is imposed such that all queries generated against a database will have preset limits corresponding to the capacity of the terminal 14.

[0061] In yet another embodiment, the truncator 152 may adjust the field size by division or other means. For example, if the display limit has been reached, the truncator 125 may reduce the field size, X by a specified amount. In an embodiment, X may be divided by two. Alternatively, X may be multiplied by a number less than 1, such as ¾, for example. Adjusting the field size allows the search engine 125 to perform more focused searches and provides more accurate search results.

[0062] In another embodiment, the truncator first attempts to display information without truncation. If that is not appropriate, the truncator may attempt truncation by beginning with one character (26 letters and perhaps 10 digits) and incrementing to two characters and then three, four, until a failure to display is reached.

[0063] In still another embodiment, the user may select a limit that will cause the truncator 152 to adjust the field size. For example, the user could specify that a maximum of ten entries should be displayed.

[0064] For certain data fields, a terminal of a hand-held device, may have a very limited display capacity. For example, a personal data assistant (POA—see FIG. 52) or a cellular phone (see FIG. 50) may be used to search a database, with the results displayed on a small screen. Alternatively a user may specify a limit on the number of entries for display. In the illustrated cases, the search engine 125 may return a result list 175 of the request 114 on multiple display pages, and the user may toggle between these multiple display pages. As an example, if the terminal 14 is limited to displaying a maximum of ten entries, and if the request 114 results in a return of a data field comprising the 400 largest cities in the United States, the truncator 152 will produce a list of 23 entries comprising 23 alphabetical characters (no cities that begin with Q, Y or Z—see FIG. 4). The search engine 125 may then display the results on three pages. Alternatively, the truncator 152 could produce a list of letter groups into which the cities would fall, such as A-D, E-G, H-M, N-R, and R-X, for example. In another alternative, the search engine 125 may send a notice to the terminal that the request 114 cannot be accommodated on the terminal 14 and may prompt the user to add an additional constraint to the request 114, so that a search result may be displayed at the terminal 14.

[0065] Adjusting the data field size also provides more convenient search results for the user. For example, if a user were to access an Internet-based database for books for sale, and were to request a list of all book titles beginning with the letter “F,” a common search engine might return several hundred titles or more, displaying perhaps twenty titles (entries) at a time. The user would then have to look through each of many pages to find a desired title. This process could be very time-consuming and expensive. Furthermore, if the search results were too large, the common search engine might return a notice saying the results were too large for display and might prompt the user to select an alternative search request. However, performing the same search using the search engine 125 allows the truncator 152 to reduce the size of the information displayed to a manageable level. In this example, if the request 114 includes the constraint “F,” the truncator 152 will loop through the data in a data field that includes book titles starting with the letter “F” until a list is available that can fit within the display limits of the terminal 14, or that fits within a limit set by the user, for example. The first list returned to the terminal 14 as a result of this request 114 may be a two letter combination with “F”as the first letter and a second letter of a book title as the second letter. For example, the fist list may include the entries “Fa,” “Fe,” “Fi,” “Fo,” and “Fu,” all of which represent titles of books. The user could then select one of the entries “Fa,” “Fe,” “Fi,” “Fo,” and “Fu” to perform a further search, continuing the process until one or more desired titles are displayed. An example of a similar truncation result is shown in FIG. 14.

[0066] When a parameter related to the search results is adequately truncated, the parameter is directed to the dispatcher 154, which retrieves the data from database 12 using the database driver 170. The dispatcher 154 then directs the final, truncated search results 175 back to the terminal 14 as a response to the request 114.

[0067] FIG. 7 is a block diagram showing the status control 140, which is responsible for monitoring the status of a current search. Due to the nature of the search engine 125, the user can choose any combination of constraints, fields or keywords, including those from past and current search cycles. The status control 140 may keep track of all past cycles of the search, as well as all information necessary to return to any of those past search cycles. The status control 140 includes a status data module 142, and an index module 144. The status data module 142 contains data related to each such search cycle, including the constraint(s) entered during the search cycle, any truncation steps taken, and the results of such truncation, for example. The index module 144 provides access to these data. When the request 114 is being analyzed by the request analyzer 130, the constraint collator 136 sends a request status query 116 to the index module 144. The status data module 142 contains information related to all past and current search cycles, which are referenced by the index module 144, and delivers a status response 119 for the most recent search cycle to the constraint collator 136. When a new constraint is sent to the query generator 150, the status data module 142 is updated 118 by the constraint collator 136. Specific structures of the request 114, the request status query 116, the status response 119 and the request status control 118 will be provided later.

[0068] The status data module 142 may be reset by the database qualifier 160 with all available fields when a refresh function is used. In an embodiment, the refresh function may be used to clear all past search cycles and the current search cycle from the status control 140. Tn such an event, the search results, such as the search results shown in FIG. 4, will no longer be displayed at the terminal 14, and data related to the past and the current search cycles may not be used for future search cycles. In effect, the refresh function may cause the entire search to be discarded. The refresh function may be activated when a user selects a refresh button (see FIG. 4) on a displayed result list, or on another portion of a GUI. Alternatively, the refresh function may discard selected search cycles. In this alternative embodiment, the user may, for example, move a cursor to a desired result list from a past search cycle and activate a refresh, reset, back, or drop button. All data associated with search cycles subsequent to the selected search cycle, including all displayed result lists may then be discarded.

[0069] FIG. 8 is a block diagram showing the database qualifier 160. The database qualifier 160 provides data field information at the start of a search or when the search engine 125 is refreshed. A field assessor 162 access the database 12 using the database driver 170, and identifies and accesses discrete data fields and other information in the database 12. A field converter 164 structures the data field information into a usable (searchable/sortable) structure and sends 163 the formatted data field information to the status control 140. Techniques for identifying and accessing the data fields, and for formatting the data field information are well known in the art. Such techniques are described, for example, in U.S. Pat. No. 5,222,066, Interface for Accessing Multiple Records Stored in Different File System Formats, the disclosure of which is hereby incorporated by reference.

[0070] FIG. 9 is a block diagram of the database driver 170. The database driver 170 is the universal interface with the database 12, which can be a local or a remote database.

[0071] FIG. 10 is an example of a search-on-the-fly using the search engine 125. In FIG. 10, a database 200 includes information related to a number of individuals. The information in the database 200 may be presented at the terminal 14 using a series of screens or menus 201-230. The user first accesses the database 200 and is presented with a list 201 of the information or data fields contained in the database 200. The result list 201 is generated by the field assessor 162, and is provided for display at the terminal 14 by the query generator 150. As shown in FIG. 10, a user has selected the data field “City” for display of information. However, the number of “cities” listed in the database 200 is too large to conveniently display at one time (i.e., on one page) at the terminal 14. Accordingly, the truncator 152 will loop a required number of times until an adequate display is available. In FIG. 10, the menu 203 shows the results of the truncation with only the first letter of a city name displayed.

[0072] Using the menu 203, the user has selected cities beginning with the letter “A.”The results are shown in menu 205. Now, the user elects to conduct another search-on-the-fly, by selecting the “sort-on-the-fly” option 206. The query generator 150 displays all the information fields available from the database 200, except for the information field already displayed, namely “City.” The results are displayed in menu 207. The user then elects to further search on the data field “State.” The query generator 150 returns the requested information as displayed in menu 209, listing five states by their common two-letter abbreviation. The user then chooses New York from the menu 209, and the query generator 150 returns a list of cities in New York, menu 211.

[0073] Next, the user elects to conduct another search-on-the-fly, option 212, and the query generator 150 returns only the remaining data fields for display in menu 215. From the menu 215, the user selects “Address” for the next data field to search, and the query generator 150 returns an menu 217 showing only first letters of the address. This signifies that the data field “Address” was too large to be easily displayed on the terminal 14. The user then elects to search on all addresses that begin with “C.” The query generator 150 returns a list of addresses by displaying only street names, menu 219.

[0074] The user then elects to conduct a further search-on-the-fly, option 220, and the remaining two data fields, “Name” and “Phone” are displayed as options in menu 221. The user selects name, and the query generator returns a further breakdown of the data by last name and by first name, menu 223. This process continues, with further menus being used to select a last name and a first name from the database 200. When the final selection is made, information from the database 200 related to the individual is displayed in window 230.

[0075] In the example shown in FIG. 10, the user could have refreshed the search engine 125 at any time, and the search would have recommenced at the beginning. Alternatively, the user could, by simply selecting a prior menu, such as the menu 215, have changed the course of the search. In this alternative, if the user had gone back to the menu 215 and instead of selecting “Address” selected “Phone,” then the menus 217-229 would be removed from display at the terminal 14, and the search would begin over from the point of the menu 215.

[0076] FIGS. 11-15b illustrate exemplary searches of a remote database, such as the database 13 shown in FIG. 1. The database in the illustrated example is for an Internet website 232 that sells books. The examples illustrated are based on a Barnes & Noble™ website. In FIG. 11, the user has applied the search engine 125 to the website 232 database, and the query generator 150 has returned a list 233 of data fields from which the user may select to access data from the website 232 database. The list 233, and other lists described below, may be displayed as overlays on the website 232. In the example illustrated, the user selects “Title” for the first search cycle. Because the list of titles is too large to easily display at the terminal 14, the truncator 152 loops until an alphanumeric list 234 is created. The list 234 is then returned to the terminal 14. For the next search cycle, the user selects titles that begin with the letter “C.” Again, the data field contains too many entries to conveniently display at the terminal 14, and the truncator 152 loops as appropriate until list 235 is created. The process continues with subsequent lists 236 and 237 being returned to the terminal 14.

[0077] FIGS. 12-15b illustrate alternate searches that may be completed using the website 232 database.

[0078] For the search results shown in FIGS. 11-15b, the status control 140 may iterate as follows:

[0079] Status Control Started . . .

• • [0080] Key: Title1 Option: Title Level: 1 Filter: Field: Title
  • [0081] Key: A2 Option: A Level: 2 Filter: SUBSTRING([Title],1,1)=‘A’ Field: Title
  • [0082] Key: AA3 Option: AA Level: 3 Filter: SUBSTRING([Title],1,2)=‘AA’AND SUBSTRING([Title],1,1)=‘A’ Field: Title
  • [0083] Key: F4 Option: F Level: 4 Filter: SUBSTRING([Title],1,1)=‘F’ Field: Title
  • [0084] Key: Fa5 Option: Fa Level: 5 Filter: SUBSTRING([Title],1,2)=‘Fa’AND SUBSTRING([Title],1,1)=‘F’ Field: Title
  • [0085] Key: Favo6 Option: Favo Level: 6 Filter: SUBSTRING([Title],1,4)=‘Favo’ AND SUBSTRING([Title],1,2)=‘Fa’ AND SUBSTRING([Title],1,1)=‘F’ Field: Title
  • [0086] Key: C7 Option: C Level: 7 Filter: SUBSTRING([Title],1,1)=‘C’ Field: Title
  • [0087] Key: Ce8 Option: Ce Level: 8 Filter: SUBSTRING([Title],1,2)=‘Ce’ AND SUBSTRING([Title],1,1)=‘C’ Field: Title
  • [0088] Key: Cells9 Option: Cells Level: 9 Filter: SUBSTRING([Title],1,5)=‘Cells’ AND SUBSTRING([Title],1,2)=‘Ce’ AND SUBSTRING([Title],1,1) ‘C’ Field: Title
  • [0089] Key: Cellula10 Option: Cellula Level: 10 Filter: SUBSTRING([Title],1,7)=‘Cellula’ AND SUBSTRING([Title],1,2)=‘Ce’ AND SUBSTRING([Title],1,1)=‘C’ Field: Title
  • [0090] Key: CC11 Option: CC Level: 11 Filter: SUBSTRING([Title],1,2)=‘CC’ AND SUBSTRING([Title],1,1)=‘C’ Field: Title

[0091] Status Control Terminated.

[0092] FIG. 15b shows the results for a search for a low-fat cookbook using the search engine 125 as applied to a remote database. In this example, the remote database is coupled to a Barnes & Noble web page. The first query, and resulting message strings, are illustrated by the following:

• [0093] Query Analyzer
• [0094] Message Received: ACK
• [0095] Status Control: Refresh
• [0096] Dispatcher
• [0097] Message Sent: Categories˜-˜Title˜-˜Author˜-˜ISBN˜SubTitle˜Format˜Date Published˜Stock Status˜Recommended
• [0098] Age˜Pages˜Ratings˜Price˜Retail˜Savings˜-˜Publisher
• [0099] Query Analyzer
• [0100] Message Received: CLK#0#1#Categories
• [0101] Status Control received an update:
• [0102] Key: Categories1 Option: Categories Level: 1 Filter: Field: Categories
• [0103] Query Generator
• [0104] Request is not cached, processing
• [0105] Generated Query: SELECT DISTINCT [Categories] FROM Books ORDER BY [Categories]
• [0106] Number of Matching Records: 2032
• [0107] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,82) FROM Books ORDER BY SUBSTRING([Categories],1,82)
• [0108] Number of Matching Records: 2022
• [0109] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,61) FROM Books ORDER BY SUBSTRING([Categories],1,61)
• [0110] Number of Matching Records: 1995
• [0111] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,45) FROM Books ORDER BY SUBSTRING([Categories],1,45)
• [0112] Number of Matching Records: 1751
• [0113] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,33) FROM Books ORDER BY SUBSTRING([Categories],1,33)
• [0114] Number of Matching Records: 1251
• [0115] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,24) FROM Books ORDER BY SUBSTRING([Categories],1,24)
• [0116] Number of Matching Records: 799
• [0117] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,18) FROM Books ORDER BY SUBSTRING([Categories],1,18)
• [0118] Number of Matching Records: 425
• [0119] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,13) FROM Books ORDER BY SUBSTRING([Categories],1,13)
• [0120] Number of Matching Records: 319
• [0121] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,9) FROM Books ORDER BY SUBSTRING([Categories],1,9)
• [0122] Number of Matching Records: 147
• [0123] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,8) FROM Books ORDER BY SUBSTRING([Categories],1,8)
• [0124] Number of Matching Records: 111
• [0125] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,7) FROM Books ORDER BY SUBSTRING([Categories],1,7)
• [0126] Number of Matching Records: 78
• [0127] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,6) FROM Books ORDER BY SUBSTRING([Categories],1,6)
• [0128] Number of Matching Records: 44
• [0129] Generated Query: SELECT DISTINCT SUBSTRING([Categories],1,5) FROM Books ORDER BY SUBSTRING([Categories],1,5)
• [0130] Number of Matching Records: 26
• [0131] Truncator finished, took 15 seconds to make 13 iterations
• [0132] Caching this request . . .
• [0133] Dispatcher
• [0134] Message Sent: Afric˜Art,
• [0135] ˜Bio˜Busin˜Compu˜Cooki˜Engin˜Enter˜Ficti˜Histo˜Home˜Horro˜Kids!˜Law:
• [0136] ˜Medic˜Mind,˜Nonfi˜Paren˜Poetr˜Refer˜Relig˜Scien˜Small˜Sport˜Trave˜Write˜
• [0137] Query Analyzer
• [0138] Message Received: CLKCategories

[0139] In the example illustrated by FIG. 15b and the above-listed message strings, an initial request would have returned 2032 book titles for cook books. This number of entries may be too large. Accordingly, the truncator 152, through 13 iterations, reduces the entries in a result list to 26. The entries in the truncated result list can then be easily reviewed by the user, and further searches may be performed to identify a desired book. As can be seen in FIG. 15b, the user has selected “Categories” as a data field to search. As is also shown in FIG. 15b, the search engine 125 may display other information windows, such as book availability, ordering and shipping information windows. With a simple drag-and-drop cursor operation, for example, the user may then order and pay for the desired book.

[0140] FIG. 16-20 are flow charts illustrating operations of the search engine 125. FIG. 16 is a flowchart of an overall search routine 250. The process starts in block 251. The request analyzer 130 receives the request 114, block 252. The request 114 may be made using a hierarchical menu-based display or a graphical user interface, with one or more layers. Using either the menu or the GUI, the user may enter specific details by typing, selection of iconic symbols or pre-formatted text, and by using well-known data entry techniques, for example. The request 114 may also comprise a simple text or voice query. Use of voice recognition may be particularly useful in mobile environments, and to speed access to the database 12. Use of voice recognition may include simple commands, such as UP, DOWN, and SELECT, to select search terms from a pre-formatted list that is presented to the user at the terminal 14. More sophisticated use of voice recognition may include actually speaking letters or numbers, or full search terms, such as speaking a key word for a key word search, for example.

[0141] The protocol analyzer 133 provides an output 135 to the constraint collator 136, and the constraint collator 136 determines the nature of the request, block 254. If the request 114 is a refresh request (i.e., a command to initiate the refresh function), the constraint collator 136 sends a reset command 131 to the database qualifier 160. The updated request 115 (possibly with a new constraint) is then sent to the query analyzer 150 for further processing, including analyzing the database 12, retrieving field descriptors, and formatting, block 256. The result of the data field descriptor retrieval and formatting are shown as an available data fields result list, block 258, and is returned to the terminal 14, block 260.

[0142] In block 254, if the request 114 is not a refresh request, the constraint collator 136 provides the updated request 115 (which may be an initial request, or a subsequent request) to the query generator 150, block 264. The constraint collator 136 compares the request 114 against information stored in the status control 140. In particular, the constraint collator 136 sends the request status control signal 118 to the status control 140 and receives the request status response 119. The constraint collator 136 then compares the request status response 119 to constraint information provided with the request 114 to determine if the constraint status should be updated (e.g., because the request 114 includes a new constraint). If the constraint status should be updated, the constraint collator 136 calls create new constraint subroutine 270, and creates new constraints.

[0143] The create new constraints subroutine 270 is shown as a flowchart in FIG. 17. The subroutine starts at 272. In block 274, the constraint collator 136 determines if the request is for a sort-on-the-fly operation. If sort-on-the-fly has been selected, field assessor 162 prepares a new set of data fields, block 280. The new set of data fields are then sent to the query generator 150, block 284, and the subroutine 270 ends, block 286.

[0144] In block 274, if sort-on-the-fly was not selected, the request analyzer 130 may receive a key word constraint, block 276. The query generator 150 will then generate an input window in which the user may enter a desired key word, block 282. Alternatively, the query generator 150 may prompt the user to enter a key word using voice recognition techniques, or any other way of entering data. The process then moves to block 284. In block 276, if a key word search option was not selected, the constraint collator 136 enters the new constraint to the existing list of constraints, block 278. The process then moves to block 284.

[0145] Returning to FIG. 16, the constraint collator 136 next updates the status control 140, block 290. In block 292, using the updated constraints, the query generator 150 generates a next query of the database 12, block 292. The database driver 170 then extracts the result list from the database 12, according to the latest query, block 294. In block 296, the truncator 152 determines if the result list may be displayed at the terminal 14. If the result list cannot be displayed, the process moves to block 298, and a truncation routine is executed. The process then returns to block 294. If the result list in block 296 is small enough, the result list is provided by the dispatcher 154 to the terminal 14, block 258.

[0146] As noted above, the request analyzer 130 determines the nature of the request, including any special commands. A special command may include a command to conduct a search-on-the-fly. Alternatively, the search engine 125 may adopt a search-on-the-fly mechanism as a default value. The search engine 125 also may incorporate other special search commands, such as a Boolean search, for example.

[0147] FIGS. 18-20 are flowcharts illustrating alternate truncation subroutines 298. In FIG. 18, the subroutine 298 adjusts a size of a data field by decrementing a parameter TP related to entries in a selected data field. For example, if the data field comprises a list of U.S. cities by name, the parameter TP may be the number of alphabetical characters in a name. The results of such a truncation is shown in the example of FIG. 4. The subroutine 298 starts at block 301. In block 303, the parameter TP is set to equal a size of the data field being searched. The truncator 152 then determines the list of records sized by the parameter TP, block 305. In block 307, the truncator 152 determines if the result list can be displayed at the terminal 14. If the result list cannot be displayed at the terminal 14, the truncator 152 decrements the parameter TP, block 309. Processing then returns to block 305, and the truncator 152 gets a reduced result list using the truncated parameter TP. If the result list can be displayed at the terminal 14, the process moves to block 311 and the subroutine 298 ends.

[0148] FIG. 19 is a flowchart illustrating an alternate truncation routine 298′. The process starts in block 313. In block 315, the truncator 152 sets the parameter TP to a size of the data field being searched. In block 317, the truncator 152 determines the list of records sized by the parameter TP. In block 319, the truncator 152 determines if the result list can be displayed at the terminal 14. If the result list cannot be displayed, the truncator 152 adjusts the size of the data field by dividing the parameter TP by a set amount, for example, by dividing the parameter TP by two, block 321. Processing then returns to block 317, and repeats. If the result list can be displayed at the terminal 14, the process moves to block 323 and the subroutine 298′ ends.

[0149] FIG. 20 shows yet another alternative truncation subroutine 298″ The process starts in block 325. In block 327, the truncator 152 sets the parameter TP to equal the size of the data field being searched. In block 329, the truncator 152 determines the list of records sized by the parameter TP. The truncator 152 then determines if the result list can be displayed at the terminal 14, block 331. If the result list cannot be displayed at the terminal 14, the truncator 152 determines if the parameter TP is less then ten, block 333. If the parameter TP is not less than ten, the truncator 152 adjusts the parameter TP by multiplying the parameter TP by a number less than one, block 337. In an embodiment, the number may be ¾. The process then returns to block 329 and repeats. In block 333, if the value of the parameter TP is less than ten, the truncator 152 decrements the parameter TP by one, block 335. Processing then returns to block 329 and repeats. In block 331, if the list can be displayed at the terminal 14, the process moves to block 339 and the subroutine 298″ ends.

[0150] The examples illustrated in FIGS. 18-20 are but a few examples of the truncations subroutine. One of ordinary skill in the art could conceive of other methods to adjust the field size. In addition to using a truncation subroutine, the user may specify a limit for the field size.

[0151] As noted above, the search engine 125 may be used for multiple searches and may be used to search multiple databases, including databases with different schemas. The results of individual searches, including the control data provided in the status control 140, are saved. The search engine 125 may then be used to further sort (search), or otherwise operate on, the results of these multiple searches. In an embodiment, the search engine 125 may perform a Boolean AND operation on two search results. The result of the Boolean AND operation would be a list of records, or entries, that are common to the two search results. FIG. 21 illustrates such a Boolean AND operation.

[0152] In FIG. 21, a GUI 400 displays local database selections 410, including a database of recordings (compact discs—CDs) 412 and a database of contacts 414. The databases 412 and 414 may be shown by text descriptions and an appropriate icon, for example. The database selections in this example are resident on a user's terminal, such as the terminal 14 shown in FIG. 1. Also displayed on the GUI 400 is a remote database selection 420 that represents databases, such as the databases 13 and 15 shown in FIG. 1, that are located remotely from the terminal 14. In the example shown in FIG. 21, the remote database selection 420 includes a database 422 for online record sales, which is represented by an icon (a CD) and a text title of the online retailer. The remote databases shown in the remote database selection 420 may include those databases for which the user has already established a link. In the example shown, the user may already have entered an Internet address for the online retailer. In addition to any returned web pages from the online retailer, the terminal 14 may then display a representation of the database 422.

[0153] Continuing with the example, the user may use the search engine 125 to conduct a search-on-the-fly of the recordings database 412 and the Virgin Records™ database 422. The user may search both databases 412 and 422 for titles of recordings that are classified as “blues.” The search engine 125 may return search results 416 and 424 for searches of both databases 412 and 422, respectively. The search results 416 and 424 may be displayed in a window section 430 of the GUI 400. The results 416 and 424 may also be represented by CD icons, such as the icons 432 and 434. The search results 416 and 424 may be stored as lists in one or more temporary databases, as represented by the windows 417 and 427. The search results 416 and 424 may also be stored in a scratch pad database 418. At this point, the user may wish to determine which recordings from the list 424 are contained in the list 416. The search engine may support this function by performing a Boolean AND operation of the lists 416 and 424. The results of the Boolean AND operation are represented by the icon 436 displayed in the window 430. To execute the Boolean AND operation, the user may simply drag the icon 432 over the icon 434, and then select AND from a pop-up menu 438 that appears when the icons 432 and 434 intersect. Other techniques to execute the Boolean AND (or another Boolean function) may include typing in a command in a window, using voice recognition techniques, and other methods. In addition, other Boolean functions may be used.

[0154] The result represented by the icon 436 of the Boolean AND operation may then be stored in a database at the terminal 14, such as in the scratch pad database 418 or may be stored at another location. The result may then be subjected to further search-on-the-fly operations.

[0155] Also shown in FIG. 21 is an online-purchase module 435 that may be used to consummate purchase of a product referenced in an online database such as the database 422. To initiate such a purchase, the user may drag an iconic or text representation of a desired product listed in the search result 424 over an icon 436 in the online-purchase module 435. This drag-and-drop overlaying these icon may initiate and complete the online purchase for the desired product.

[0156] Use of the search engine 125 may be facilitated by one or more GUIs that are displayed on the terminal 14. FIGS. 22-26 are examples of such GUIs. In FIG. 22, a GUI 450 includes a display section 452 and one or more database sections such as local database section 470 and remote database section 460. The local database section 470 includes databases local to the terminal 14. In the example shown, the local databases include a patients database 472, a general contacts database 474, a pharmacy database 476, a medicines database 478 and a scratch pad database 480. The remote databases include an Amazon.com database 462, an online record retailer database 464, a Physician's Desk Reference database 466 and an American Medical Association (AMA) online database 468. The remote and local databases may be represented by a text title and an icon, both contained in a small window as shown. A user may access one of the remote or local databases by moving a cursor over the desired window and then selecting the database. In the example shown, the local medicines database 478 has been selected, and a list 490 of data fields in the medicines database 478 is displayed in the display section 452. Also included on the display section 452 is a keyword button 492 that may be used to initiate a key word search of the medicines database 478.

[0157] FIG. 23 shows the GUI 450 with a user selection of a category data field from the list 490. The category data field is indicated as selected by an arrow adjacent to the data field name. When the category data field is selected, a category list 494 is displayed on display section 452. The category list 494 includes four entries, as shown.

[0158] The user may continue to search the medicines database 478 using key word techniques and search-on-the-fly techniques. FIG. 24 shows the GUI 450 with results of several search cycles displayed.

[0159] FIG. 25 illustrates a search of the PDR database 466. Such a search may be initiated by dragging a cursor to the window having the PDR 466 symbol (text or icon), and then operating a “select” button. FIG. 26 shows a search of the Amazon database 462. This search may also be initiated by a “drag-and-drop” operation.

[0160] The SOTF search engine 125 may accommodate merging of one or more sets of search results. The multiple search results may be derived from a common database, or from more than one database. A search using the search engine 125 may be controlled through a user interface by one or more icons that can represent (1) filters or (2) the images of filters. Thus, the icon may represent spatial or temporal attributes, or sets of objects or procedures. Merging the icons thus has two interpretations corresponding to (1) and (2): either filters are added (“apply every filter in every icon to every image to which it can be applied”), or image sets are added. In an alternative embodiment, the addition (union or join) operator may be any other relational operator, e.g. divide, difference.

[0161] Use of the merge function may be explained by reference to Boolean lattices. A collection of entities can have attributes A or B or both. If {A} is the set of all A entities and {B} is the set of all B entities; the set whose elements all possess both attributes A and B may now be written {A and B}, and the set whose elements all possess either attribute A or attribute B or both may be written {A or B}. The elements of {A and B} can be considered to possess a new, less inclusive or specific attribute C, and the elements of {A or B} to possess a new, more inclusive or general attribute D. In a lattice, the nodes are attributes; the most inclusive attribute (in this case D) is always at the top and is called the join of those attributes (nodes) immediately below it, and the most exclusive attribute (in this case C) is always at the bottom and is called the meet of those attributes (nodes) immediately above it. In other words, the OR operation yields the join of two attributes, while the AND operation yields their meet. Thus, the OR operator is upward or inductive (yielding the more inclusive join of the operands), while the AND operation is downward or deductive (yielding the more exclusive meet of the operands). The nodal attributes of such a lattice are analogous to filters; but since a principle called CF duality states that attributes and sets are to some extent interchangeable because every attribute characterizes a set and every set is characterized by an attribute, these attributes are logically equivalent to the sets they characterize.

[0162] In an example optical context, the downward AND operator corresponds to stacking colored filters, while the upward OR operator corresponds to mixing colored paints or filters. In color optics, stacking and unstacking colored lenses is called a subtractive process, while mixing or unmixing paints is called an additive process. Unfortunately, while combining or “adding” filters is subtractive with respect to the sets they characterize, it is additive with respect to the filters themselves, and adding sets is subtractive with respect to the filters. So it is better to refer to operations among attributes (filters, lenses, etc.) as “filtrative” or “infonegative, and to those among sets (paints, lights, etc.) as “constructive” or “infopositive”. CF duality can now be rephrased as follows: every infonegative entity (attribute) descriptively characterizes an associated infopositive entity (set/object), and every infopositive entity instantiates or is descriptively characterized by an associated infonegative entity.

[0163] The search engine 125 may include iconization (iconic representation) of an algebra or calculus of relations defined on Boolean lattices. This representation begins with a set of primitive icons extracted from base tables and defines new icons (derived tables, virtual databases) by means of simple user-executed operations. The icons can be effortlessly translated into lists of data corresponding to the icons, and it is these lists that comprise the real substance of any search procedure.

[0164] When search chains are branched into to chains A and B, the filters subsequently applied to each chain can be the same or different, and merging can signify any of two or more Boolean relationships (relational operations) defined on a relational database. Specifically, when chains merge, sets of filters can be added or intersected. Since filters are constraints, adding them amounts to intersecting their images, while adding their images amounts to intersecting the filters (infopositive-infonegative distinction). Equivalently, one may consider positive and negative filters effecting deduction and induction respectively; the filters are descriptive, while the images are substantive. The extent to which the images of filters can intersect depends on the commonality (predicative non-exclusivity) of domains. Icon algebras (of iconic operators) are “object-oriented”on the GUI level; they are Ul extensions of the innate object-orientation of relational databases themselves, wherein the objects are records, attributes, tables, virtual databases and so on, and the operations are those of any relational algebra.

[0165] The looping and merging of search chains is to some extent algebraic. First, since actual topology is being changed, such transformations do not directly form a topological homeomorphisin group; the algebra remains Boolean, and the “homeomorphism” is defined on the operator graph of the Boolean algebra (of which the initial search tree is generally only a subspace). Icons representing sets of nested predicates are “Boolean objects”; when decision chains converge or diverge, objects merge or split, and these objects represent (combinatorially) unique search paths. Thus, operations among paths can be reduced to operations among objects; e.g., regress-diverge is just an object-splitting operation. Continuous looping applies “inverse deductive filters” to achieve induction by descriptive intersection of filter constraints, permitting the retrograde convergence of paths to identical ancestral objects (inductive merging of objects), while inductive looping is just direct regression to an ancestral object preparatory to splitting it and thus effecting divergence of paths (deductive splitting of objects). Deductive convergence of paths is “natural” if iconic image sets intersect and “forced” if not; if natural, then there has been non-exclusivity of subobjects, and paths are not unique (even though identical filters can apply to divergent paths without impairing uniqueness). So all deductive merging is forced, and this entails a decision regarding which filters are to be conserved and which discarded. Any such operation will effectively “rewrite the paths”, and doing this optimally is NP-complete.

[0166] More specifically, icons are subject to CF duality. The merge control thus has a “switch” toggling between “Qualities/Objects”. When the switch is in the “qualities” position, merging icons performs a qualified deductive conjunction of filters and yields a set intersect; when it is in the “objects” position, merging the icons performs a disjunction of filters and an inductive union of sets, yielding a more general attribute (the general qualities created by the object-merge operation will be produced by sets of filters applied disjunctively). The search engine 125 is therefore capable of inductive and deductive information processing. A quality-merge in which filters do not cross the line between composite icons equates to an object merge; the set thus selected is characterized by a more general quality which amounts to the descriptive (filtrative) union. There is also a modified quality-merge in which filters in either icon applicable to both iconized sets are applied to both, thus crossing the line between icons. In this case, a true merging of paths occurs, as opposed to path icons. The search engine 125 allows users to choose which filters are to cross the inter-icon line and which are not, resulting in complex Boolean expressions and the sets they characterize (determining consistency of complex expressions can amount to LSAT; sets of inconsistent expressions will simply yield a null return.

[0167] Icons may reside in the first menu box to appear, being transferred from menu to menu as the path is generated and filters are accumulated. When a direct regress occurs, the path is regarded as “complete” and is stored in a holding module. Prior to the merging operation, the quality/object switch is set; and icon subfilters or subsets individually displayed. A “lattice navigator” will keep track of position and equivalence, folding the search graph in case a node of the original tree is inductively encountered in the course of an object-merge; otherwise, the icon remains in “internodal space” (which is to be regarded as a virtual space realized only in the event that the search tree is nondisjunctive in its nodes and therefore incomplete with respect to the semantic net generated by the tree).

[0168] FIG. 27 is a flow chart illustrating an alternative operation 600 of the query generator 150 of FIG. 6. In the illustrated operation, the query generator 150 is adapted to receive multiple selections of items within a same menu function and within a same merge function. To provide this functionality of the query generatory150, the request analyzer 130 (see FIG. 5) may be adapted to receive a collection of user choices.

[0169] The operation 600 begins in block 601. In block 603, the request analyzer 130 receives constraints collected from the constraint collator 136, and the updated request 115, which may be an initial request or a subsequent request, is provided to the query generator 150. In block 605, the query generator 150 determines if the constraints (the request 115) are in the same merge group. If the query generator 150 determines that the request 115 is in the same merge group, the process moves to block 607 and the query generator 150 generates the query with a Boolean AND. If the request is not in the same merge group, the query generator 150 generates the query with a Boolean OR, block 609.

[0170] In block 611, the items selected within the same unit are Or'ed and the default truncator may be used depending on the size of the returned items. In block 613, the generated query is executed. In block 615, the number of records to be displayed is checked. If the number is within a specified limit, the process moves to block 617 and the search results are returned for display. The operation 600 then ends, block 625. In block 625, if the number of records to be displayed is too large, the process moves to block 621, and a truncation routine is executed.

[0171] The truncation routine may be any of the previously-described truncation routines illustrated in FIGS. 18-20. FIG. 28a illustrates an alternate truncation routine 630. The routine 630 begins in block 631 with the truncator 152 receiving the request 115. In block 633, the truncation is set to the size of the field being viewed on the GUI, and sets the False Flag. The query is then run against the database using the selected truncator, block 635. In block 635, the truncator 152 determines if the number of records that would be retrieved from the database can be displayed on the existing GUI. If the records can be displayed, the process moves to block 639, and the truncator 152 determines if the Flag is set False. If the Flag is set False, the process moves to block 653 and the records are returned (displayed on the GUI). The process then ends, block 655. In block 637, if the number of records exceeds the display size of the GUI, the status of the Flag is checked as False. If false, the truncator is set to 1, and the flag is set to true, block 647, and the process returns to block 635. If in block 637. If the flag is not set false, the process moves to block 651, and saved records are retrieved. The retrieved records are then displayed, block 653.

[0172] In block 639, if the Flag is not set to false, the retrieved records are saved, and the truncator 152 is incremented. The process then returns to block 635.

[0173] FIG. 28b illustrates another alternative truncation routine 700. In block 701, the truncator 152 receives the constraints, the view by field and the maximum of number of display items (MNDI). In block 702, the truncation is set to zero (no trunction), and the Flag is set to True. Next, the query is generated in block 702. In block 703a, query generator receives the constraints, the view by field, and the truncator as parameters, and the query generator returns the query. The query is then run against the database, and the counter is set to zero, block 704. In block 705, the truncator 152 fetches the next record and increments the counter. If the end of file is reached, block 706, and the truncation equals zero, block 710, the truncator 152 sends the list of fields to the client, block 712. However, if the truncation is not zero, block 710, the truncator 152 is incremented, block 709, and the process returns to block 703. On the other hand, if the end of file is not reached, block 706, and the counter is smaller than MNDI, block 707, the process goes back to block 705, in which the truncator 152 fetches the next record and increments the counter. However, if the counter is larger than MNDI, block 707, and the saved list of fields exist, block 708, the truncator sends the list to the client, block 712. Conversely, if the saved list of fields do not exist, block 708, the truncator 152 is incremented, block 709, and the process goes back to block 703 again.

[0174] Table 1 illustrates an example of the alternate truncation routine 700. This routine begins by attempting not to truncate the records.

[0175]

	TABLE 1
	Records	1st Round	2nd Round	3rd Round
1	Armandia	1	Armandia	1	A	1	AR
2	Armonk	2	Armonk	2	N	2	NE
3	Armonk	3	New Orleans	3	R	3	RI
4	New Orleans	4	New York			4	RO
5	New Orleans
6	New York
7	New York
8	New York
9	Riverdale
10	Riverdale
11	Riverdale
12	Rockfort

[0176] In this example, the maximum number (n) of displayable results is three, and the database contains twelve instances of six different cities. First, the database is queried for the full city field with no truncation, and records are fetched. Records are fetched until four (n+1) records are fetched from the database. Since the number of different cities (4) is greater than n, fetching is halted and the process moves to truncation. Then the database is queried for only the first letter of the cities (truncation is incremented so that it equals one). For this query the database manager may simply review its index. The compiled list from the query is saved as “A”, “N”, and “R”. Next, the database is queried for the first two letters of the city field (truncation is incremented so that it equals two). Again, the database manager may simply review its index to locate this information about the data field. This query for two letters or characters is continued until the number of two letter combinations exceeds n. When the number of different combinations (4) is again greater than n, the routine halts and nothing is saved. The system now returns to the previous saved list. Therefore, the saved list (“A”, “N”, and “R”) is returned to the client for display or process.

[0177] FIGS. 29-38 illustrate graphical user interfaces and search on the fly results using the search engine 125 with a merge function. In FIG. 29, a search of a patent database has been executed to search for patents by primary examiner. The Primary Examiner results table lists the arabic numerals 0-7 and the letters A-Z, indicating that the database contains names of primary examiners beginning with these numerals/letters. To quickly narrow the search, the user selects the letter O, and results are returned listing last and first names all primary examiners whose last name begins with O. As can be seen by the returned results, the database lists several primary examiner instances of O'Dea. This could indicate an error in the database. The search engine 125 allows these errors to be detected and corrected. The correction may be made by selecting the incorrect instances, right-clicking the correct instance, and then choosing a “correct all other's based on this instance” function.

[0178] FIG. 30 shows how multiple-select capabilities of the search engine 125 may be used to enhance a search. In the illustrated example, the user searches for 3M Company. Different versions of the company name are then displayed with the returned results. In this way, the user may select the different versions of the company that the user wants to use for the search. The pop-up pane shows a current status control for the GUI.

[0179] FIG. 31 shows the results of subsequent menus showing the aggregation, or merge, of two previous constraints, “3m” and 3-M.” FIG. 32 shows a merge execution. The user first selects the “3-M” and the “3M” company names using the check boxes in the previous menu. The user then selects the merge option, placing the menu on hold, and going to the “M”, “MI”, “MIN” and “MINNESOTA M” menus. The merge option is then selected on the menu and the merged menu is displayed showing the merge of searches between “3M” and “Minnesota Mining and Manufacturing Co.” FIGS. 32-36 show other search engine 125 features including data mining and database correction.

[0180] FIGS. 37-39 show the results of a full text search of a patent database using the keyword “encryption” and searching on all fields. The initial search results are truncated to display by first letter/numeral of the patent title. From this intermediate search result menu, the user selects all patents whose title begins with the letter “E”, and a subsequent search result menu is displayed listing partial titles of all such patents. From the next intermediate list, the user selects the patent whose title begins “Electronic copy protection mechanis” (see FIG. 38) The search engine 125 then returns this specific patent, the first page of which is shown in FIG. 39. The displayed patent includes the keyword “encryption” highlighted wherever it occurs. The display also indicates the number of instances of the keyword in the patent.

[0181] FIGS. 40-49 illustrates additional search results.

[0182] In the examples shown in FIGS. 37-49, search results are displayed on a “large-format”screen, such as available with a desktop personal computer. When a user is in a mobile environment (e.g., on foot, in a car) the user may still be able to access the search-on-the-fly search engine and have search results returned to a mobile display device such as a cellular telephone or a personal data assistant.

[0183] FIG. 50 illustrates a standard cellular telephone 800 that may use the search-on-the-fly search engine 125. The cellular telephone 800 includes a display 801, a keypad 802, and other controls 803 that may be used to navigate one or more data buses using the search-on-the-fly search engine 125.

[0184] FIG. 51 illustrates a personal data assistant (PDA) 800 that may use the search-on-the-fly search-on-the-fly search engine 125. The PDA 800 includes a display area 811 and an input area 812.

[0185] FIGS. 52a-52l illustrate a search sequence using the cellular telephone 800 configured to use the search-on-the-fly search engine 125. In the example illustrated, the U.S. Patent and Trademark Office patent database is selected. Using the cellular telephone 800, the user conducts a search of the U.S. Patent and Trademark Office database using a series of filters. Each time a filter is applied, a search result may be returned and displayed on the display 801. Using the controls 802, the user may add or subtract filters. The display 801 shows the accumulative result of the filtering process. When the data to be returned is too large to fit the display 801, the returned data may be truncated as illustrated in FIGS. 52f-52k.

[0186] FIG. 53 illustrates a general purpose personal computer system 850 that may be used for search-on-the-fly of a plurality of databases. The system 850 includes a processor section 851, a display and a control section coupled to the processor section 851, and a computer readable medium 855, which may be read by components of the processor section 851. The computer readable medium 855 may include the software routine required to implement the search-on-the-fly with merge function method.

[0187] In specific embodiments, the search engine 125 is implemented as a program executed on a general purpose computer, such as a personal computer. The search engine may also be implemented as a routine attached to a database structure. In addition, the search engine may be implemented on any processor capable of executing the routines of the program. In alternative embodiments, the search engine 125 may be implemented as a single special purpose integrated circuit (e.g., ASIC) having a main or central processor section for overall, system level control, and separate circuits dedicated to performing various different specific functions, computations and other processes under control of the central processor section. Those of ordinary skill in the art will appreciate that the search engine 125 may also be implemented using a plurality of separated dedicated or programmable integrated circuits, or other electronic circuits or devices (e.g., hardwired electronic or logic circuits such as discrete elements circuits, or programmable logic devices, such as PLDs, PLAs, or PALs). In general, any device or assembly of devices on which a finite state machine capable of implementing flowcharts similar to the flowcharts of FIGS. 16-20, 27 and 28 can be used to implement the search engine 125.

[0188] While using search on the fly has been described in detail for an end result of printing, viewing or displaying data, search on the fly can be useful for other purposes. Search on the fly does not require obtaining the underlying data in the database or the display of the underlying data to be useful. Search on the fly can be used for gathering information or characteristics about data in a database with or without downloading the data itself. This gathered information about the data can be used to analyze the data, sorting, correct or clean data, verifications and confirmations. For example, search on the fly can be used to determine whether there is existing data in a database within certain ranges or parameters (date ranges, numerical, alphanumerical and other characteristics). If there is data within certain parameters, the number of datapoints within those parameters can also be determined. This information about the data can be gathered using search on the fly with queries to the database manager (which may simply need to query its index and not access the data itself). Another example is correcting data. Data may need to be corrected or cleaned for various reasons including spelling errors. Search on the fly can locate these errors without necessarily accessing and downloading the data itself. Certain combinations of characters or truncations will be obvious spelling errors. Also, data that is out of range can be located and corrected or eliminated from the database using search on the fly. Another example is data from one database can be confirmed or verified against data in a second database using search on the fly. Those skilled in the art will find many uses and specific applications for search on the fly.

[0189] The terms and descriptions used herein are set forth by way of illustration only and are not meant as limitations. Those skilled in the art will recognize that many variations are possible within the spirit and scope of the invention as defined in the following claims, and there equivalents, in which all terms are to be understood in their broadest possible sense unless otherwise indicated.