Generating and searching compressed data6937171Abstract Data destined for a client is compressed at a server in a manner that produces a compressed data string that can be searched in its compressed state. The server constructs a code table that assigns codes from a standard code set (e.g., ASCII code set) that are normally unused to selected character pairs in the data string (e.g., the most frequently occurring character pairs). During compression, the selected character pairs are replaced with the corresponding codes. Identifiers are inserted into the compressed data string to separate substrings. To search the compressed data string at the client, a search query is compressed and compared to the compressed substrings. The substring identifiers are used to quickly locate each successive compressed substring. When a match is found, the matching substring is decompressed by replacing the code in the compressed substring with the corresponding character pair in the code table. Claims 1. A system for preparing program data for delivery to a client that executes an electronic program guide, comprising: Description TECHNICAL FIELD
Then, for the smaller memory section 404(5) associated with Friday, the data selector 224 might select only the following collection of EPG data:
The EPG data included in Friday's memory section 404(5) is sufficient for the viewer to browse the future programming and to set various conditions, such as reminders or recording events. FIG. 5 shows one example of a selection process 500 performed by the time-based data selection program 224 executing at the EPG server 110. Generally, the process 500 can be thought of as two passes over the time-divided memory structure to dynamically store as much EPG data in each of the memory sections. The first pass begins at the day furthest in the future and works backwards toward the present day. The second pass starts with the current day and works forward into the future. In both passes, unused space from one day is shifted to the next day. The process 500 may be implemented in software, firmware, hardware, or a combination of these. The process is illustrated as a set of operations that may be implemented as computer-executable instructions that can be executed by a computer, such as EPG server 10. At block 502, the process allocates disproportionate amounts of the memory 402 (FIG. 4) among the time units. For discussion purposes, suppose the total amount of memory available for EPG data is 500 K bytes and the process allocates this memory over five 24-hours time slots as represented in FIG. 4. As an initial allocation, suppose the section 404(1) for Monday is allocated 120 K bytes, the section 404(2) for Tuesday is allocated 110 K bytes, the section 404(3) for Wednesday is allocated 100 K bytes, the section 404(4) for Thursday is allocated 90 K bytes, and the section 404(5) for Friday is allocated 80 K bytes. At block 504, the first or future-to-present pass is initiated, starting with the section corresponding to the furthest time in the future for which there is EPG data. Here, there are five days of program listings and the fifth or furthest day out from Monday is Friday. At block 506, the process determines whether all of the EPG data for Friday will fit in the allocated space 404(5) (e.g., 80 K bytes). If there is more EPG data than available space (i.e., the "No" branch from block 506), the process removes one or more items of EPG data (block 508) and reevaluates whether the reduced data set fits in the allocated space. This loop is repeated until a set of EPG data that fits the space is found. The removal of EPG data may be handled in a number of ways. For instance, the EPG data may be prioritized in some manner that establishes the order in which items are removed. For the EPG data shown in FIG. 4, one possible order might be as follows:
Another possible approach is to assign level-of-detail values that correspond to diminishing sets of EPG data. For instance, a first value might represent the fewest number of acceptable items in the EPG data for a give time unit; another value might represent the next biggest set of EPG data; and so one. With this approach, the operation of block 508 is to find the appropriate level-of-detail value for the available memory space. Once the appropriate amount of EPG data is found to fit the allocated space for Friday (i.e., the "yes" branch from block 506), the EPG data (or corresponding level-of-detail value) is temporarily stored memory (block 510). At block 512, the process determines whether all allocated time units in the first pass have been considered. In this case, only Friday has been evaluated. Thus, at block 514, the process continues to the next time unit, which is Thursday in our example. Additionally, the process may optionally add any available space into the allocation for the next day to be considered. For instance, suppose that the first iteration found that 78 K bytes of EPG data fit the allocated 80 K bytes. The additional 2 K bytes would then be shifted to the memory section 404(4) for Thursday to enlarge that allocation to 92 K bytes of memory. The first pass through the available memory continues for each day, from Friday back to Monday. With each iteration, any additional memory space is moved to the next day to accept potentially more EPG data for that day. Once the EPG data set for the current time unit (i.e., Monday) is found and recorded at block 510, the process begins the second pass at block 516. This present-to-future pass begins with the next closest time unit (i.e., Tuesday). Any leftover space that was not used to hold the EPG data for Monday is added back to the available space for Tuesday to see if any additional data can now be fit into Tuesday's allocated space (block 518). At block 520, the process determines whether any more EPG data for Tuesday will fit in the allocated space. With the additional space moved over from Monday, there may be just enough space to add one or more EPG data items or modify the level-of-detail value. If more EPG data can be added, the process enlarges the EPG data for that day (block 522) and reevaluates with the enlarged data set whether even more data may be included. This loop is repeated until a set of EPG data that fits the space is found. Once the appropriate amount of EPG data is found to fit the allocated space for Tuesday (i.e., the "no" branch from block 520), the data items and/or corresponding level-of-detail value associated with this day is updated to reflect the additional subject matter (block 524). At block 526, the process determines whether all allocated time units in the second pass have been considered. If not, the next iteration is performed for the next time unit (e.g., Wednesday) and any available space is moved into the allocation for that time unit. The second pass through the available memory continues for each day, from Tuesday back to Friday. When all time units are considered (i.e., the "yes" branch from block 526), the EPG data file is constructed (block 528). The resulting file includes varying amounts of data with more EPG data being contained in the current day and progressively less EPG data for subsequent days. The space used to store the various days of EPG data will most likely be different from the initially allocated amounts as a result of the two-pass process. As an example, the final distribution of available memory space might be 123K for Monday's EPG data, 109K for Tuesday's EPG data, 104K for Wednesday's EPG data, 88K for Thursday's EPG data, and 76K for Friday's EPG data. In any event, the EPG data is guaranteed to fit within the pre-allocated space at the client (e.g., the 500 K bytes of memory). Table Arrangement of Sorted EPG Data Low resource client 130 has limited memory resources (e.g., 500 K bytes) and limited processing resources to perform operations on the data, such as searching. Accordingly, one process performed on the EPG data prior to delivery to the client concerns structuring the EPG data in a way that facilitates searching at the client. The EPG data is pre-sorted at the EPG server 110 according to data type, such as by titles of programs. In one example, the EPG server 110 can be used to pre-sort those items of EPG data selected as a result of the time-based selection process described above with respect to FIGS. 4 and 5. The pre-sorted EPG data is arranged in tables that are delivered and stored at the client. The pre-sorted tables are conducive to fast searches at the client, even though the client has limited processing capabilities. It is noted that the sorting process may be performed at other computing sites in system 100, including at the head end services 120. FIG. 6 shows exemplary EPG data 600 to be delivered to the client. The EPG data 600 is stored in multiple tables, where individual tables are employed to store similar data of a particular type. In the illustrated example, there are one or more program tables 602(1), 602(2), 602(3), . . . , 602(p) containing program listings, one or more schedule tables 604(1), . . . , 604(s) containing schedule-elated data, a station table 606 containing station information, and a channel table 608 containing real/virtual channel and frequency information. The tables store the program data in records. For example, the program tables 602 consist of records pertaining to programming information, as represented by program record 620. Each program record 620 has one or more fields, such as a program identifier field 622, a program title field 624, a program description field 626, and so on. The schedule table 604 has records pertaining to scheduling information, as represented by program record 628. Each schedule record 628 has one or more fields, such as a time field 630 and a program identifier field 632. The station table 606 has records pertaining to station information, as represented by station record 640. These station records 640 may include one or more fields used to identify a station, such as its call letters, a corporate entity description, and so on. The channel table 608 stores records pertaining to channel information, as represented by channel record 650. The channel records 650 may include one or more fields used to identify a channel, such as a channel identifier, a channel frequency, and so on. The tables are related so that records in one table can cross-index into related information in one or more other tables. For instance, the channel record 650 in channel table 608 may index to a station record 640 in the station table to identify the call letters for the channel identifier. That station record 640 may then index into the schedule table 604 to identify multiple records 628 defining the schedule for that station. The schedule records 628 may further reference individual program records 620 in the program tables 602. The program records 620 hold the details of the programs corresponding to the time slots identified in the schedule table 604 for the particular station or channel. The tabular data structure 600 represents the native form of the EPG data that can be delivered, via broadcast or other means, to the clients. The EPG data can be indexed by channel to allow individual networks to obtain the appropriate EPG data for their clients. The EPG program at the client understands the native form and can process the data to perform various operations. In its unsorted state, however, the EPG data is not conducive to fast searches on the low resource client. The client is faced with either sorting the existing EPG data on its own, which is resource expensive, or performing lengthy searches on unsorted data, which is slow. FIG. 7 illustrates one exemplary implementation of the sorting process 700 for sorting the EPG data 600. The process is described with reference to a diagrammatic illustration depicting exemplary components of the television entertainment system 100 that might be used to perform the enumerated operations. At 702, the program data is stored in native form in multiple tables. This is represented by the tabular data structure 600 shown in FIG. 4, and diagrammatically illustrated in FIG. 5. The program listings are thus originally 18 stored in program tables 602 in the order produced by the EPG data publisher 102. At 704, the data sorter 222 at the EPG server 110 sorts the program data in the tables by a specific parameter type that a viewer is anticipated to want to search. For example, in one implementation, the data sorter 222 arranges EPG data in the program tables alphabetically according to the "stopped name" of the program. The "stopped name" of a program is the shortened version of the program title that contains the identifying words, without common joiner words such as "the", "and", etc. For example, the movie "The Good, The Bad, and The Ugly" might have a stopped name of "Good, Bad, Ugly" and the program "How the West was Won" might have a stopped name of "West Won". An example set of program records 620 for different programs is shown in FIG. 7. Notice that the program record for the movie "The Good, The Bad, and The Ugly" is ordered alphabetically under "G" rather than under "T". Similarly, the movie "The Matrix" is arranged under "M" and the program "How the West was Won" is sorted under "W". The data may be sorted using other data types as well. For example, the data sorter 222 may arrange the EPG data according to actor names, or program genre, or ratings. If there is EPG data for multiple days, the data sorter 222 sorts the program records for each day independently of the other. This will produce multiple sets of sorted program data for corresponding multiple days. Alternatively, all of the records for all of the days can be sorted together. It is also noted that other servers besides the EPG server 110 may be employed to perform the sorting techniques as described herein, such as servers at head end services 120. At 706, the EPG server 110 constructs a data file 750 for delivery to the client. The data file 750 holds the tables, including the sorted program tables. If there is EPG data for multiple days, one data structure is constructed for each day. The one or more data structures 750 may then be broken into multiple chunks that may or may not be the same size. Each chunk is atomic, containing a range of records and an index to describe where the records in the table(s) fit within the context of the whole EPG data file. Each chunk may hold parts of one table or up to multiple tables. Individual tables know in which chunk they reside. Table boundaries are therefore different than chunk boundaries, although they may coincide. At 708, the data file 750 is delivered to the client 130 via the head end service. When the client 130 receives the data file, it stores the data file in RAM 310. At 710, the client performs a search over the data structure stored in RAM. Due to the pre-sorted arrangement of the records, the client is able to perform a simple and fast binary search on the data. For instance, suppose viewer is interested in locating the movie, "The Matrix". The client's search engine 324 performs a binary search through the title fields of the program records, comparing two titles at a time. Here, the movie "The Matrix" is ordered after "Mary Poppins" and before "Miracle on 34th Street" as represented by title listing 752. Where data structures for multiple days exist, the search engine 324 performs a two-phase searching process. A first phase involves a binary search of the program data for each day to produce intermediate results. Suppose, for example, that the program data covers five days, Monday through Friday. The first phase produces search results for each data, Monday through Friday. A second phase combines the daily results produced from the first phase and sorts them. The final results are then returned. Fragmentation of Program Data Another process that may be performed on the EPG data prior to delivery to the client concerns formatting the EPG data into multiple blocks of a predetermined size. The client 130 designates an arbitrary data size and allocates a portion of its memory in segments of that size. The arbitrary size is communicated to the EPG server 110. The data structure fragmenter 220 "fragments" the publisher-created EPG data 104 stored in the EPG database 108, or a subset of that data, in advance of delivery to the client 130. When finally delivered, the fragmented data fits neatly into the pre-allocated segments of the client memory. As a result, system calls to the memory for purposes of accessing EPG data are greatly reduced or eliminated, resulting in less fragmentation of memory and more efficient storage of the EPG data. It is noted that the fragmentation process may be performed at other computing sites in system 100, including at the head end services 120. FIG. 8 illustrates one exemplary implementation of the fragmentation process 800. The process is described with reference to a diagrammatic illustration depicting exemplary components of the television entertainment system 100 that might be used to perform the enumerated operations. At 802, the client 130 designates an arbitrary data size and allocates a portion of its memory 310 into segments 820 of that size. As one example, the memory segments 820 are equal size segments of 16 K bytes. This memory allocation size might alternatively be specified by the manufacturer. At 804, the client 130 communicates the size of the memory segments 820 to an upstream server, such as EPG server 110. Alternatively, the segment size of the client memory may be a parameter that is already known to the EPG server 110. For instance, the clients may be configured during manufacturing to allocate the memory designated for holding EPG data in certain size segments. This parameter could then be provided to the EPG server 110 as the target size for the data fragments. It is also noted that other servers besides the EPG server 110 may be employed to perform the fragmentation techniques as described herein, such as servers at head end services 120. At 806, the EPG server 110 begins producing the EPG data file for delivery to the client. The EPG server 110 may structure and format the data file in many ways. One approach is to construct multiple tables that hold various types of EPG data, such as the table structure 600 of FIG. 6. Each table is self-contained in that it knows its type and contents. The tables are arranged in a data structure, which is represented as structure 822 in FIG. 8. The table data structure 822 is broken into multiple chunks 824 that may or may not be the same size. Each chunk 824 is atomic, containing a range of records and an index to describe where the records in the table(s) fit within the context of the whole EPG data file. Each chunk 824 may hold parts of one table or up to multiple tables. Individual tables know in which chunk they reside. Table boundaries are therefore different than chunk boundaries, although they may coincide. At 808, the data structure fragmenter 220 fragments the table structure 822 into smaller size fragments 830. Each structure fragment 830 is capable of fitting in a corresponding memory segment 820. More particularly, in our example, the EPG data fragments 830 are of a size that is guaranteed to be less than or equal to the arbitrary size designated by the client 130, or less than or equal to 16 K bytes. Notice that the fragment boundaries may or may not coincide with the chunk boundaries and/or table boundaries. At 810, the fragmented data file 114 can be delivered to the client 130 via the head end service. When the client 130 receives the fragmented data, the client stores the data fragments 830 in respective pre-allocated segments 820 without making system calls to the memory. When the client subsequently makes a call to free memory, the memory is provided in the fixed-size segments. In this way, fragmentation is constrained to the fixed-size spaces that can be more easily managed, thereby eliminating the need for memory management techniques such as de-fragmentation or garbage collection processes. It is noted that the fragmentation process may be performed on either non-compressed or compressed data. If performed on compressed data, the fragmentation process is executed following the compression process described in the next section. Generating Searchable Compressed Data Due to the limited memory at the low resource client, another process that can be performed on the EPG data prior to delivery is to compress the EPG data in a manner that facilitates searching of the data in its compressed state. The following discussion provides one exemplary approach to generating and searching compressed data. The techniques described below can be performed on essentially any string of bits, and is particularly well suited for alphanumeric text strings. For continuity purposes and ease of understanding, the techniques are described in the context of compressing the EPG data at the EPG server and then searching the compressed EPG data at the client. The general process is described first, followed by a more detailed explanation of one exemplary implementation that utilizes the ASCII standard. FIG. 9 shows an exemplary process 900 for generating and searching compressed data. Generally, the process 900 includes a server-based phase and a client-based phase. In this server-based phase, data (e.g., EPG data) is compressed into a format that is readily searchable by the client, including low-resource clients. This first phase can be performed by the data compressor 226 executing at the EPG server 110, or by other computing sites in system 100, such as at the head end services 120. The client-based phase is performed at the client. It involves searching the compressed data in its compressed state, as well as decompressing the data when used. The process 900 may be implemented in software, firmware, hardware, or a combination of these. The process is illustrated as a set of operations that may be implemented as computer-executable instructions that can be executed at the server and the client. A dashed line distinguishes the operations being performed in the two phases by the different computing devices. At 902, the process receives an input data string and makes a pass through, counting each pair of characters. This data string can be essentially any string of alphanumeric characters. In our example, the data string is a string of primarily text characters that form the EPG data in the fragmented tables. The data compressor 226 constructs a counts table that contains entries for every possible character pair, and counts the number of occurrences of each pair in the input data string. When the entire data string has been evaluated, the data compressor ascertains which character pairs are the most common by comparing the counts. Those pairs with the highest counts are the most common. At 904, the data compressor 226 constructs a code table that associates codes with characters and the most common character pairs found in the data string. The code table contains a standard code set (e.g., ASCII code set) that includes codes for every character in the data string as well as codes that are normally unused. The data compressor 226 assigns unused codes to the most common character pairs identified in the counts table. At 906, the data compressor 226 compresses the alphanumeric data string using the codes in the code table. The most common character pairs are replaced with shorter codes to reduce the amount of data. This substitution produces a compressed data string that is significantly smaller than the original input string. At 908, to facilitate searching of the compressed data string, identifiers are inserted between substrings to separate the compressed data string into searchable portions. For instance, in the EPG data, the process may place identifiers (e.g., one or more zeros) between program titles, or actor names, or any other items that a viewer is anticipated to want to search. At 910, the server packages the compressed data string and code table in a data file that is delivered to the client. The data file may be directly distributed to the client, or via the head end services. At 912, the client stores the compressed data string and code table in RAM 310, as shown in FIG. 3. This is beneficial for low-resource clients with limited memory capabilities because more EPG data can be stored at the client. At 914, the client is able to search the compressed data string in its compressed state. In the described implementation, the search engine 324 searches the compressed EPG data by compressing at least a portion of the search query and then comparing the compressed search query with substrings in the compressed data string. The search engine can rapidly skip from substring to substring by keying on the identifiers that separate the substrings. At 916, the client-side decompressor 320 can decompress all or a portion of the compressed data string using the code table. The decompressor 320 passes through the compressed data string, substituting the character pairs in the code table for the associated codes in the compressed string to expand the data string back to its original size. Various operations of the process 900 will now be described in more detail. For discussion purposes, the detailed example is described in the context of using the 256-character set of ASCII (American Standard Code for Information Interchange) codes. Each ASCII character is represented as an eight-bit byte, which can be represented in hexadecimal as a set of codes ranging from 00 (i.e., 0000 0000) to ff (i.e., 1111 1111). It is noted, however, that other code sets may be used, such as EBCDIC (Extended Binary Coded Decimal Interchange Code), UTF8, 8-bit Unicode, and the like. Table Construction (902 and 904) FIG. 10 shows one exemplary implementation of the table construction operations of 902 and 904 in FIG. 9. The process is described with reference to a diagrammatic illustration depicting a counts table 1100 and a character code table 1200. At 1002, the counts table 1100 and the character code table 1200 are initialized. An exemplary counts table 1100 is shown in more detail in FIG. 11. It is a 256×256 table. The rows represent the first character in a character pair of the input data string and the columns represent the second character in the character pair. Thus, the counts table 1100 contains entries for every possible character pair combination. An exemplary character code table 1200 is shown in more detail in FIG. 12. It is a 256×2 table, with 256 rows representing the 256 ASCII codes and two columns. Since not all codes are used to represent single characters and other punctuations (e.g., commas, periods, quotations, etc.) in the data string, some normally unused codes are available to represent the most common character pairs identified in the counts table 1100. In the described implementation, the tables 1100 and 1200 are initialized to all zeros. At 1004 in FIG. 10, the data compressor 226 makes a first pass through the input data string and counts each occurrence of character pairs. With each count of a character pair, a corresponding entry in the counts table 1100 is incremented. Suppose the input data string contains the textual substring " . . . the next big thing . . . ", identified by reference number 1050. As the compressor 226 evaluates this substring 1050, it increments the count for the character pairs "th", "he", "e-", "-n", "ne", and so on (where "-" represents a space). Notice in FIG. 11 that a count field 1102 associated with the character pair "th" is incremented to a value "1" as shown. The character "t" is represented in ASCII as hexadecimal 74 (decimal 116) and the character "h" is represented in ASCII as hexadecimal 68 (decimal 104). Similarly, a count field 1104 associated with the character pair "he" is incremented to a value "1". As the data string is traversed, the count values are incremented to reflect the number of occurrences of individual character pairs. At 1006, the data compressor determines the N most common character pairs in the data string. Character pairs are ascertained as being the most common by comparing the counts. Those pairs with the highest counts are the most common. The number N is a variable that can be preset or made dependent upon how many codes in the code table 1200 are available for assignment to character pairs. At 1008, the data compressor 226 marks the single characters found in the input data string as used in the code table 1200. When a character is found in the data string, the value in the right column "R" is changed to 1 to represent that it is being used. This is represented in FIG. 12 with entries for "t", "h", and "e". At 1010, after all single characters and any punctuation symbols are marked, unused codes in the code table 1200 are assigned to represent the most common character pairs with the highest counts. As shown in FIG. 12, the left column "L" holds the first character and the right column "R" holds the second character. Suppose that the character pairs "th" and "t-" are found to be among the most common character pairs. An unused entry 8f (or 143) in code table 1200 is assigned to character combination "th" and an unused entry bd (or 189) is assigned to character combination "t-". Compression (906 and 908) FIG. 13 shows one exemplary implementation of the compression operations of 906 and 908 in FIG. 9. The process is described with reference to a diagrammatic illustration of progressively compressed strings. At 1302, every character pair in the input data string is evaluated. A portion of our example data string is shown as " . . . the next big thing. The story about . . . ", and is referenced as number 1350. At 1304, if a character pair is represented in the code table (i.e., one of the most common character pairs), the code is substituted for the character pair. In the illustrated example, the character pair "th" is twice replaced with the code value 143, and the character pair "t-" is once replaced with the code value 189. This produces a compressed data string, as represented by compressed portion 1352. Each substitution results in a 50% reduction as the two bytes of each character pair are replaced with a one-byte code. It is noted that one or more passes may be made through the data string to produce the compressed data string. Notice also that the character pair "Th" is different than the character pair "th" due to the different case of the letter "T" in the former, and hence is not replaced with the code for "th". Additionally, the single characters in the compressed data string are shown in their character format for ease of discussion and illustration, but may be replaced with their corresponding codes in the compressed data string. At 1306, compressed substrings are separated with a unique identifier. In this example, a zero is used to separate adjoining substrings, as represented in compressed data string portion 1354. However, identifiers other than zero may be used. This separation facilitates quick searching of the compressed substrings, even at a low resource client. At 1308, a compressed data string with identifier-separated substrings is output. Client-Based Searching (914) FIG. 14 shows one exemplary implementation of the searching operation 914 in FIG. 9 that is performed at the client on the compressed data. The process is described with reference to a diagrammatic illustration of an example search item. At 1402, the client receives a search item. This search query may be generated, for example, when the viewer selects a program from the EPG user interface. The search item can be on essentially any type of data. In the context of EPG data, the search item may be a title, actor name, rating, program genre, schedule time, station, and so forth. An example search item 1450 contains the text string "the next big thing." At 1404, the client compresses the search item using the same compression process described with reference to FIG. 13. The client examines each character pair in the search item and replaces pairs represented in the code table with the code. This produces a compressed search item 1452. At 1406, the client indexes to the first or next substring in the compressed data string stored in memory. The client uses the substring identifier (e.g., a zero) to rapidly skip from substring to substring. At 1408, the compressed search item is compared to at least a portion of each compressed substring. In this example, the compressed search item 1452 is compared with each compressed substring 1354. When no match is found, the client proceeds to the next substring by skipping ahead to the next identifier. If a match is found, the client decompressor 320 decompresses the substring and any related substrings at 1410 to reproduce the original string 1350. Client-Based Decompression (916) FIG. 15 shows one exemplary implementation of the decompression operation 916 in FIG. 9 that is performed at the client. The process is described with reference to a diagrammatic illustration of decompressing the compressed substring 1354. At 1502, the decompressor receives the compressed string of data, as represented by substring 1354. The decompression operation may be performed on the entire data string, or portions thereof. At 1504, the decompressor 320 makes a pass through the compressed data string. When the decompressor encounters a code, the decompressor uses the code to index the code table 1200 and replaces the code in the compressed string with the associated character pair from the code table 1200. This is illustrated by the code 143 referencing the associated character pair "th" in the code table. When the string portion "143e" is decompressed, the code 143 is replaced with the character pair "th" to thereby expand the data string back to its original content. At 1506, the decompressor 320 outputs the decompressed data string, or substring, for use by the client. In the context of EPG data, the decompressed data string can be passed to the EPG 322. CONCLUSION Although the invention has been described in language specific to structural features and/or methodological acts, it is to be understood that the invention defined in the appended claims is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are disclosed as exemplary forms of implementing the claimed invention.
|
Same subclass Same class Consider this |
||||||||||