TRS(R) Enterprise Database Server 6.80 Copyright 1993-2012, Beijing TRS Information Technology Co., Ltd. ================================================================= TABLE OF CONTENTS ================================================================= A. Contact Technical Support. B. Startup and Shutdown. C. Install and Upgrade. D. Requirements. E. What's new? ================================================================= A. Contact Technical Support ----------------------------------------------------------------- If you have any technical problems, and can't settle them, please contact with our Technical Support as followings. Address: Beijing TRS Information Technology Co., Ltd, 17/F, Tower A, Team Center, Science Garden, Datun Road, Chaoyang District, Beijing P.R.C., 100101 Web: http://www.trs.com.cn, http://www.trs.cn Mail: support@trs.com.cn, info@trs.com.cn Tel: 86-10-64848899 Fax: 86-10-64889088 ================================================================= B. Startup and Shutdown ----------------------------------------------------------------- 1. Run trsserver(.exe) to start up TRS Server. 2. Run trssearcher(.exe) to start up TRS Searcher. 3. Run trsmirror(.exe) to start up TRS Mirror. 4. Run tdbshutdown(.exe) to shutdown TRS Server and TRS Searcher. 5. Run tdbshutdown(.exe) with option '-cache' to merely shutdown TRS Searcher. 6. Run trsmirror(.exe) with option '-shutdown' to close down the TRS Mirror. Notes: 1. In Microsoft Windows NT, you can start and stop the Server at the Control Panel of service. 2. A service port number(or name) can follow the trsserver(.exe) /trssearcher(.exe)/trsmirror(.exe). e.g: trsserver 8888. 3. A service port number (or name) and option '-kill' can follow the tdbshutdown(.exe). e.g: tdbshutdown [-kill] 8888. Notes: The option '-kill' of tdbshutdown maybe unsafe for TRS Server, thus you do not use it unless no other ways to close. ================================================================= C. Install and Upgrade ----------------------------------------------------------------- 1. Run tdbinstall(.exe) to install product or demo system. 2. Run tdbupdate(.exe) to upgrage system. Notes: 1. In UNIX, you must create a user account for TRS at first, and then run the tdbinstall program under this user or the supper user (recommended) of system. 2. You should feedback the displayed hostID of the machine to us to produce a product KEY for your installation need. 3. The tdbupdate program can update the system data dictionaries and TRS system files automatically. So you should place those files together with it. And before starting tdbupdate command, you must shutdown TRS System. ================================================================= D. Requirements ----------------------------------------------------------------- 1. UNIX, or Linux, or WindowsNT 4.0 with Service Pack 3 or above. 2. 1024 MegaBytes RAM recommended at least. 3. Approximate 1024 MegaBytes free disk space. 4. 1024 MegaBytes swap space recommended. Notes: You should prepare enough additional free disk to accommodate the growing databases while you load data into TRS System. ================================================================= E. What's new? ----------------------------------------------------------------- New in TRS 6.80(BUILD:5290, 2012/7/6): 1. Support parallel indexing to improve system performance while creating index, loading records and copying records. 2. Enhance the XML processing to support a XML_KEEP_CDATA option in the trs-parser instruction. 3. Add two simple relevance methods (HIT_TF and LENGTH) for some special applications. 4. The tdbshutdown supports '-reboot' option, and tbdmaster adds the '-pathset' command. 5. Improve LIKE function's performance and effectiveness. New in TRS 6.80(BUILD:5220, 2011/4/11): 1. Support logout the batch of user instances to reduce network traffic for the cluster application environment. 2. Support super quote marks for Javabeans API to construct the parameters of retrieval function. 3. Support the extension A (GB/T20542-2006) and the extension B (GB/T 22238-2008) of Tibetan coded character set. 4. Add a interface to support deleting a database from the view. 5. Improve the limit strategy of the column value size so as to be able to store larger files. 6. Support the hit point summary to enhance the overview effect. 7. Support time-weighted to improve the effect of relevance for the news databases. 8. Improve index, the comma is no longer a sentence punctuation. 9. Support a retrieval option to ignore all the punctuations to improve the response performance. 10. Support new functions: restore the records logically deleted, and move the database to other directory. 11. The tdbmaster supports to backup and restore(show) databases. 12. Calculate more weightings to improve the effect of relevance ranking for the retrieval results. 13. Support people's name and entity's name recognition for LIKE and INCLUDE functions to improve the effect of relevance. 14. Support complex parameter of INCLUDE function to improve the effect of relevance ranking. New in TRS 6.80(BUILD:5030, 2010/5/12): 1. Support a type of new index structure for better performance of relevance ranking. 2. Support quickly searching, so it's possible to return a part of results before finishing the whole computing. 3. Support two kinds of new sorting methods for Chinese text in databases, including PINYIN and BIHUA. 4. Support more factors for relevance computing: MULTI_RELEVANT, HeadingWeight, KnowAssWeight, and BranchWeight. Consider the number of hit values for CHAR columns with multi-values. 6. Function LIKE supports flexible and self-adjusted similarity for the retrieval results by Top N sorting. 7. Add a parameter for function INCLUDE to support the phrase's segmentation by user with space character. 8. Support the limitative operation with COL to limit computing in the same column for the application's complex condition. New in TRS 6.50(BUILD:4140, 2010/4/12): 1. Support expansion searching of SYNONYM/ANTONYM/THESAURUS for CHAR column. 2. Support different weight values for the ontic columns in the dummy logical column. 3. Support the expansion searching of cognate words for english. 4. Support locating of source codes for system error. New in TRS 6.50(BUILD:4130, 2009/12/28): 1. Support almost all the letter-words (include Greek) to index. 2. Support to protect the Database's backup for Runtime Edition. 3. Support to encrypt and compress the downloaded record's file in TRANSPORT format for Runtime Edition. New in TRS 6.50(BUILD:4100, 2009/9/18): 1. Add system log for starting or shutdown server. 2. Add function to modify register code in tdbmaster(.exe) tool. New in TRS 6.50(BUILD:4090, 2009/9/2): 1. Support word's list while reading information of index words. New in TRS 6.50(BUILD:4080, 2009/7/6): 1. Support the different timeouts with every searching requests. New in TRS 6.50(BUILD:4070, 2009/5/6): 1. Support word's segmentation interface with default lexicon of system charset for client's applications. New in TRS 6.50(BUILD:4050, 2009/4/30): 1. Upgrade the scheduler module of service to improve efficiency and allow more client's connection cache. 2. Upgrade the chinese word's segmentation algorithm and lexicon to decrease the ambiguous and redundant phrases. And allow to load vocabularies by not lexicon's character set but client's local character set. 3. Improve log module to allow saving by day and recording error. 4. Support redundant index for CHAR column. 5. Support updating batch records with different column's values in different records, and adding new records at the same time. 6. Support the dummy logical column based on BIT column. 7. Improve the RANGE retrieval function to support searching the numerical value within specified range in text. 8. Improve the LIKE retrieval function to enable counting result words of segmentation to debar the single ideograph character. 9. Support the SPELL and NEAR retrieval functions. 10. Support sentence tag in text, and disable normal punctuations. 11. Support calculating the relevance by cosine formula. 12. Support dual machine's ID codes for new license KEY. New in TRS 6.10(BUILD:3390, 2008/10/28): 1. Improve sorting of PHRASE column by ignoring the prefix LF/CR characters in column's value. New in TRS 6.10(BUILD:3383, 2008/9/4): 1. Support showing the basic information of users, databases and columns with tdbmaster. New in TRS 6.10(BUILD:3382, 2008/8/7): 1. Support client's stable accessing for Wide Area Network users. New in TRS 6.10(BUILD:3380, 2008/7/16): 1. Support multiple limiting for one element of Native XML Index in retrieval expression. New in TRS 6.10(BUILD:3370, 2008/7/7): 1. Improve process of pruned TOP N sorting with limit of records. 2. Optimize the performance of fuzzy searching without Hit Point for PHRASE and DOCUMENT columns. 3. Allow ignoring error report for inexact columns while getting result records. New in TRS 6.10(BUILD:3362, 2008/6/10): 1. Update Quick Sort algorithm to adapt the extreme bad data set. 2. Optimize the performance of Unique Value Checking for loading records and creating indices. 3. Support pre-accessing some database's files after starting up to enable the cache of file system. New in TRS 6.10(BUILD:3361, 2008/4/25): 1. Support local Throught Access (source files) mode for loading records. New in TRS 6.10(BUILD:3360, 2008/4/23): 1. Support the TRS TRANSPORT format file with inline binary data while loading records to avoid creating large number of affix files for TRS STANDARD format. New in TRS 6.10(BUILD:3303, 2008/4/3): 1. Enhance system's ability to correct damaged data dictionaries. New in TRS 6.10(BUILD:3300, 2008/2/29): 1. Support the TOP N sorting and pruned TOP N sorting methods to improve performance for sorting the very large result set. 2. Support classifing function of column value to output classes with record-number order and store partial classes only. 3. Support pre-loading the column's value allowed cache and norm length of text column allowed relevance sorting. 4. Support setting condition of record number to spilt databases for each automatic-mode VIEW. New in TRS 6.10(BUILD:3211, 2008/2/14): 1. Improve the SYNONYM and ANTONYM dictionaries to permit a word appearing multiple groups. 2. Improve regular expression to support Chinese with [\CH] rule. 3. Improve Cluster IP list of Background running mode to support wildcard matching, and single native IP address (127.0.0.1 or one of others). New in TRS 6.10(BUILD:3210, 2008/1/16): 1. Support ignoring the errors reported by OEM document's filter. 2. Support Ignoring the punctuations of sentence, if they appear in the phrase of expression searched. 3. Support classing the DATE/NUMBER column's data with a regular expression. And improve regular expression to support Chinese code or Unicode. New in TRS 6.10(BUILD:3200, 2007/12/18): 1. Support the file cache for retrorsely reading data to improve the performance of relevance sorting by multiple text columns when the file system cache is invalid. 2. Support relevance sorting without regard to the word's IDF to avoid the different relevant degree result of same data among the different databases. New in TRS 6.10(BUILD:3195, 2007/11/1): 1. Support home directory with space characters from the command line input while installing and upgrading. New in TRS 6.10(BUILD:3194, 2007/10/29): 1. Support bitmap to speed up the process of optimizing database. 2. Implement Optimizing/Checking/Indexing databases functions in the tdbmaster. New in TRS 6.10(BUILD:3191, 2007/9/27): 1. Support outputing partial results for word-selection function. 2. Support outputing the Hit Points when a cut-size is specified. 3. Improve process of lexical parsing to deal with the very very long text and large numbers of words. New in TRS 6.10(BUILD:3190, 2007/9/18): 1. Improve the user rights management of automatic-mode VIEW for uploading records or inserting record. 2. Improve the performance for maintaining database's indices on Windows platform by asynchronously closing the index files. New in TRS 6.10(BUILD:3183, 2007/9/1): 1. Enhance the ability of data recovery when optimizing database. New in TRS 6.10(BUILD:3181, 2007/8/8): 1. Improve the socket's transmission performance with TCP option TCP_NODELAY turned on for the UNIX/Linux platform. New in TRS 6.10(BUILD:3180, 2007/8/2): 1. Support Windows x64 platform (just recompile). 2. Support loading share libraries dynamically in the 32 bits OS patform of AIX with 4.3 version (just recompile). New in TRS 6.10(BUILD:3160, 2007/7/25): 1. Improve the report of valid records after classing the result set by column's values. New in TRS 6.10(BUILD:3150, 2007/7/23): 1. Upgrade the gcc compiler in Linux platform to avoid some bugs about the O2 optimization option. TRS Server generated by old compiler does not work well when searching POS4 index. New in TRS 6.10(BUILD:3100, 2007/4/16): 1. Support storing the values of columns concentratedly that are allowed to put into cache, and the normalized lengths of TEXT columns which are allowed to sort by relevance. 2. Support sorting result records by a relevance synthetized the weights of multiple TEXT or other columns that are allowed to participate in the sorting by relevance. 3. Support tracing the Hit Points and sorting the result records by relevance with all instead of exact positions of hit words, in other words, without saving exact Hit Points. 4. support forcing a long paragraph to split for 'SEG;POS' index. 5. Support indexing the string connected by the dot as a unitary word, and redundant indices for a internet domain name. 6. Allow inserting record and updating records with control file. 7. Support filtrating the column's value by a regular expression compatible with Perl to classify result records. 8. Support outputing DATE and NUMBER column's value with formats. 9. Improve the maintaince of SEGMENT, STOP and SPARSE lexicon to become effective instantly without restarting TRS Server. New in TRS 6.00(BUILD:3005, 2007/3/30): 1. Enhance checking the version of data and index files, and set the thread stack size automatically for UNIX when the default configuration of system is too large or small. New in TRS 6.00(BUILD:3003, 2007/2/9): 1. Support entering nice sleep while maintaining data or indices of database to give more CPU time to other processes if which are executing the requests about database retrieval. 2. Improve the Instance's management for offline-working request. 3. Improve the setting of error number when a request is aborted. 4. Improve the inside operation of LIKE function. New in TRS 6.00(BUILD:3002, 2007/2/5): 1. Support scanning the all network cards to match licensed host. 2. Improve the errors' process of child-threads in a ThreadGroup. 3. It is enhanceed to check the CryptCode when logout a Instance, and support to kill all clients. New in TRS 6.00(BUILD:3001, 2006/12/18): 1. Disable the aborting process when updating database's records. 2. Support fuzzy matching n charachers at most with a '%' symbol followed by n '?' symbols, such as 'X%????Y' (from 0 to 4). New in TRS 6.00(BUILD:3000, 2006/7/30): 1. Improve the searching performance with new algorithm of Index. New in TRS 5.00(BUILD:2003, 2006/3/15): 1. Improve the classifing and pruning functions about result set. 2. Support multiplex operation on the DOCUMENT and PHRASE column. 3. Support checking the file system whether it allows large file exceeding 1G or 2G. New in TRS 5.00(BUILD:2002, 2005/10/29): 1. Support cache of the data values for DATE/NUMBER/CHAR columns to be used by sorting, classifing, DATA function and etc. 2. Support classifing and pruning retrieval result with value of the column specified. 3. Support surrogate code pair in UTF8 charset when indexing and searching. 4. Support saving the text data only drawed from document by OEM filter when loading records. 5. Improve the DATA function to support fuzzy searching, and the RELEVANT degree to support method: half of consine value plus half of specified column's value. 6. Improve the searching with index by character but no index of single character. 7. Improve indexing and searching the long rule string connected by the charcter such as '#&/-_'. New in TRS 5.00(BUILD:2001, 2005/7/4): 1. Improve the limit mechanism of memory usage to avoid bringing temporary files as possibly to speed up the searching process. 2. Support virtual retrieval result histories. 3. Improve the LIKE function with two modes of 'AND' and 'OR' to support matching partial words, and allow using the operators such as SEG/SEN/POS to limit result set size. 4. While outputing result records, support drawing out text data automatically from formatting document data such as WORD/HTML in DOCUMENT column if no text is stored. 5. Support '/frec=n' limit operator, means that the appearing of the word or phrase limited in a record must satisfy the given condition here. 6. Improve IDLE's automatic function. Support minimal optimizing and creating indices. Support defining task rules to describe task's objects and time, and etc. 7. Support chinese number when importing data into NUMBER column. 8. Support statistic on machine load information for the dynamic feedback balance method of TRS Cluster. New in TRS 5.00(BUILD:2000, 2005/4/25): 1. Support UTF8 character set, native XML indexing and searching. 2. Support segmentation for Latin/Cyrillic/Arabic words based on the GB18030 or UTF8 charset. 3. Support automatical extension of searching between simplified Chinese and traditional Chinese based on a same database with the GB18030 or UTF8 charset. 4. Support the system native character set (includes GB18030 and BIG5) to convert text automatically from the character set to the local character set (includes GB18030, BIG5 and UTF8) set by new client applications. 5. Resolve the problem between SIG_IGN and wait calling in Linux. 6. Resolve the problem that can't report the occurred error when searching a VIEW. 7. Resolve some problems of loading database records such as the MULTI_VALUE data on non MULTI_VALUE column, and too long TEXT data for a column. 8. Support stem index for the languages: Danish, Dutch, English, Finnish, French, German, Italian, Norwegian, Russian, Spanish, Portuguese, Swedish. 9. Support creating the mixed index of CHARACTER and WORD in the same time. And allowing to ignore the high-frequency WORDs or VALUEs while searching. 10. Improve the algorithm for computing relevance weight of words. 11. Support confining a IP address set to login TRS Server by the CONNECT class user. 12. Support LIKE function to change the full text retrieval style, and DATA function to scan data directly, and $CODE to produce the natural sequence numbers as the NUMBER column values. 13. Support automatic-mode VIEW for Runtime Edition of TRS Server. 14. Support copying records from a ordinary VIEW, and sorting the records while copying. New in TRS 4.51(BUILD:1051, 2003/9/11): 1. Support GB18030 character set standard (four-bytes character). 2. Support automatic-mode VIEW to split database while inserting /coping/loading/updating/deleting records. 3. Improve the managing mechanism about the record ID filter for deleting records both physically and logically, and the cache mechanism of TRS database record's link and mirror's files. 4. Implement an embedded XML filter to index the XML format data as a column value, and support high light points in retrieval result with the specified tag and mode. 5. Allow to alter the attributes of column index before creating. 6. Support rebooting server automatically with the given periods. 7. Support displaying lincece number of TRS Server. 8. Improve 'pre/text' operator to check on same SEG and same SEN. 9. Enhance the ability to capture the abnormal running status of TRS Server for WIN32 system. 10. Don't check the validity of month's days in retrieval express. 11. While optimizing, enhance to automatically repair the damages of databases' main index file on some especial condition. 12. For a new created databases, support the rule-words connected by charcter '#', and allow to change the indexing means about those rule-words connected by the charcter such as '#&/-_' by setting the external configuration. 13. Locking up the target databases while backuping and restoring. 14. Enhance the ability to read a bad backup package of databases. 15. Improve the algorithms to speed up the outputing a lot of IDs of result records at one times. 16. Improve the method about automatically renaming database when restoring a backup of database with a long name (e.g. exceeds 30 characters). 17. Enhance the ability on verifing INVERT index file block while maintaining database's indices to avoid the potential problem begotten allocating huge size memories mistakenly. 18. Support loading data of ISO2709 into TRS database by the name or alias of column automatically, and extracting the sub-data into a specified column alone. 19. Improve the t40master to display files copied one by one when backuping and resuming system, and check the disk's available space before backuping system, and ignore the invalid paths. 20. Support sorting result records by the values of PHRASE column. 21. Resolve the problem that makes Server deadlock while multiple users (exceed three) load/insert/update/delete records on the same database and at the same time. New in TRS 4.50(BUILD:1050, 2002/5): 1. Support sorting result records by the unit vector's length of words hit to replace the relevance-sort method. 2. Support automatically aborting the executing idle-task of TRS Server when the ending condition has come. 3. Support storing the text data from office documents extracted automatically by OEM filter. 4. Enchance the ability of HTML filter to recognize the abnormal tags such as '
'. 5. Delete the databases created by backup-restoring process when the operation is failed. 6. Support the highest and lowest limit for the number of result records allowed to sort, specified by a searching option. 7. Enhance to process the iterative values of PHRASE or DOCUMENT column when loading database records. New in TRS 4.50(BUILD:1049, 2002/4): 1. Speed up searching for the fuzzy ascii string in the DOCUMENT and PHRASE columns of a Chinese database with BI-GRAM indices. 2. Enable searching a english-stop-word with a fuzzy symbol such as the 'and%' or '?and', if the 'and' is a stop-word. 3. Automatically abort the long-time requests such as optimizing database, loading records, and so on, while the TRS Server is shutdown. 4. Improve the means to judge that a file or a directory whether exists or not on WIN32 platform. New in TRS 4.50(BUILD:1047, 2002/3): 1. Enchance the ability of HTML filter for analysing error pages. New in TRS 4.50(BUILD:1045): 1. Support parallel retrieval based on mutli-threads system when the retrieval target includes multi-databases, and in mechine, there are multi-CPUs. 2. Append a DOCID to the specified sorting methods automatically to speed up the sorting process on a big result set while the value's kind of target column is very few. 3. Provide a special null SEGMENT lexicon to index whole Chinese characters under the indexing mode based on Chinese character. 4. Improve the indexing means of special ascii strings connected by the charcter such as '&/-_' for new created databases. 5. Support the dafalut_retrieval_column_logic_priority mode when some dafalut_retrieval_columns are specified. And the primary method, expression_logic_priority mode is the dafault option. 6. Support encrypting the user's password when it is changed, or a new user is created. New in TRS 4.50(BUILD:1044): 1. Improve the cache mechanism of TRS database's main index file. New in TRS 4.50(BUILD:1043): 1. Add a coefficient to the IDF of words so as to ensure that it is always bigger than 1, and enable relevance-sorting anytime. New in TRS 4.50(BUILD:1041): 1. Improve the support about XML record format. 2. Provide a run-time edition of TRS Server for OEM. 3. Support large file (64 bits file system) on 32 bits platforms. 4. In retrieval process of 64 bits platform, the index-item uses 32-binary-bit integer unit instead of 64 bits integer unit to reduce the usage of memories. New in TRS 4.50(BUILD:1040): 1. Support Safe-Interrupt mode to ensure that the database index is complete while loading/inserting/updating/deleting/copying records or creating/removing indices. 2. Support the English words connected by the character '&'. 3. Enable Multiplex-OR-Operation for 'Column_List += Expression'. 4. Support automatic extension retrieval between ASCII character and Chinese-ASCII character. 5. Support loading records written by XML into database directly. 6. Support OEM document filter and URL GET-METHOD while loading/ inserting/updating/copying records. 7. Unlimit the size of files stored with ALONE mode in database. 8. Support TAG prefix in TRS Format Record File. New in TRS 4.50(BUILD:1039): 1. Support sub-types of three kinds for NUMBER column: 32 binary bit single precision float, or 64 binary bit double precision float, or 32 decimal bit signed integer. 2. Automatically Terminating the retrieval process if TRS Server has received the SHUTDOWN command. New in TRS 4.50(BUILD:1038): 1. Improve the fuzzy searching for a bi-gram index based on word. 2. Enable to re-create index on database column if the old index file of the column can't be found in the given directory. 3. Allow to delete the FirstTextColumn column of database at any time, and to re-define it while the index of object column is yet empty or deleted before doing. 4. Check the available space of devices while loading/inserting/ updating/copying records and creating the index of column, to avoid that the devices overflow. At same time, check the file size to avoid to overflow. 5. Allow the '_' character to be the first character or the last character in TRS object name (such as TRS user name, database name, etc.). 6. Check the free space of temp-device while optimizing database. 7. Support a new API to calculate the disk space occupied by the each given database. 8. Add a new setting (BTREEINDEX) in the CACHE.CFG file to allow the B+TREE index placed into a memory cache while maintaining database indices. New in TRS 4.50(BUILD:1037): 1. Adopt the new word-segmentation system: support bi-gram means based on Chinese word and character to speed up searching for phrase; improve on searching of ascii string; and process the ambiguous Chinese strings. 2. Add the MaxConcurNum setting in the NET.CFG file to limit the maximum number of concurrent processes or threads of Server. 3. Support to backup the whole system online, and to resume when the system is down. 4. Change the method of transmiting socket between two processes in LINUX if its kernel version is above 2.2, so as to support some LINUX such as Turbo, and enable that starting TRS Server by any LINUX user is same as by the LINUX super user. 5. Enable optimizing database, if some data blocks of record are damaged on UNIX platform. New in TRS 4.10(BUILD:1036): 1. Add a option (REMOTE) for TimeWaitMust in the NET.CFG file to distinguish local connections from remote connections. New in TRS 4.10(BUILD:1035): 1. As well as possible to detect the amount of physical memories of current machine automatically while installing & upgrading TRS Server. But some platform can't be detected such as IRIX, SCO and so on. 2. Support to configure the running mode of TRS Server about the connection with client and the TCP's TIME_WAIT state. In this way, you can select the best mode as you want. New in TRS 4.10(BUILD:1034): 1. Improve the SERVICE_GET_RECORDID and SERVICE_PICK_RECORDID to unlimit the number of record IDs output at one times. 2. Allow to alter the database's (and column's) attributes about index (e.g: the SEGMENT lexicon quoted) when the index is not created yet, even if there are records in the database. 3. Improve the SEGMENT lexicon loading to accept word attributes. 4. Reset the database when all of its columns are deleted. New in TRS 4.10(BUILD:1033): 1. Support the mirror of TRS Server to synchronize the specified databases records in multiple sites. 2. Support to modify encrypting of records without updating data. 3. Support to flush the running-log file of TRS Server using the t40master command with '-flush' option. New in TRS 4.10(BUILD:1032): 1. Support the unique checking of columns when loading/inserting or updating database records. 2. Improve the searching support about '%' and '?' as a signifi- cative not wildcard character. And the fuzzy searching method of a phrase is improved too. 3. Improve the outputing format of large value of NUMBER columns. 4. Support to empty the SEGMENT/STOP/SPARSE(etc.) lexicon. New in TRS 4.10(BUILD:1031): 1. Support shared cache of phrase and express of relevant column in TRS Searching Server. 2. Improve the method of identifing a machine to try to generate a stable host ID. New in TRS 4.01(BUILD:1030): 1. Allow to insert any printable characters into STOP lexicon. 2. Support the plug-in of professional TRS Searching Server with shared cache (word only for relevant column), and multithread running mode. 3. Support to automatically cancel the multiple search-instances of the same license, and keep up the last instance only. 4. Support GBK character set for indexing and searching. Perhaps you must re-build the index of data, if you require searching a phrease including GBK character. 5. Dealing with 65535 of CUT_SIZE specially while getting result records, to return whole text data of DOCUMENT column without non-text data, and ignore all data of BIT column. New in TRS 4.01(BUILD:1029): 1. Optimize the implementation of AND class operators between an exact word and a phrase string. 2. Enable case insensitive while wandering around directories to find some things on WIN32 platform. 3. Enable optimizing database, if some data block of records are damaged. New in TRS 4.01(BUILD:1025): 1. Check the free space whether enough or not when backuping and restoring databases. 2. Allow to change some attributes of database and column, which are irrelative with loaded datas and indices. 3. Potentiate the checking of TRS object names and their aliases to avoid repeat. 4. Support huge size files with 4G bytes in 32-bit NT system. New in TRS 4.01(BUILD:1024): 1. Improve the performance with highly-frequency words retrieval. 2. Support the backuping of databases exceeded 2G bytes. 3. Implement the utility tool to diagnose and display TRS system data dictionaries. New in TRS 4.00(BUILD:1023): 1. Improve the retrieval method of a phrase string, if it is the result of SYNONYM/ANTONYM/THESAURUS functions, or is executed with automatic extending retrieval. 2. Alter the interface of SERVICE_PICK_RECORDID. New in TRS 4.00(BUILD:1022): 1. Implement the running-log of Server. 2. Implement the supervisor (t40master) of Server's status. 3. Implement the local database-loading utility (t40load) in the Server site. 4. Update the data dictionary about SEGMENT/STOP/SPARSE(etc.) to let its item of 'NumOfRecords' express the number of words in the lexicon. 5. Support canceling automatically the overdue login when Server is idle. New in TRS 4.00(BUILD:1021): 1. Implement the functions of selecting/inserting/deleting words from/into SEGMENT/STOP/SPARSE(etc.) lexicon. 2. Support searching records without retrieval histories written. 3. Support retrieval with SEN and SEG attributes in DOCUMENT and PHRASE columns even if they don't have these attributes. New in TRS 4.00(BUILD:1020): 1. Improve the date format checking, for example: the '1999/2/29' is a invalid date. New in TRS 4.00(BUILD:1018): 1. Support using $DATE(...) to express a date relatived to today, that the something omited is a number or a negative number of relative days. 2. Support using $TIME(...) to express a date relatived to today, that the something omited is a number or a negative number of relative seconds. 3. Support the storage method of index without compression. 4. Support the LIFO of physical record ID of retrieval result. 5. Support exporting record directly by physical record ID. New in TRS 4.00(BUILD:1017): 1. Improve the method of coping records, to overcome the problem that need a good deal of disk space when the number of target records is very big. 2. Log the motion track of updated records (if moved), to assure that the agone retrieval result set is valid, and the logical ID of updated record is unaltered. 3. Improve the method of updating records, which distinguish the PHRASE and DOCUMENT type of target columns from other types. New in TRS 4.00(BUILD:1015): 1. Support the '$CODE' as an automatic value of CHAR column when uploading records into database, and the each automatic value is unique and invariable and ordered in whole database. New in TRS 4.00(BUILD:1014): 1. Differentiate the default storage path between the dictionary and database. 2. Support multiple segmentation, user definition and stop-words dictionaries in one Server. 3. Support both BIG5 and GB Chinese character sets in one Server. New in TRS 4.00(BUILD:1013): 1. Limit the max length of record for inserting/updating/reading. 2. Support Chinese BIG5 character set. New in TRS 4.00(BUILD:1012): 1. Support changing TRS object name between upper and lower case of english letter. 2. Enable the digit indexing attribute of database when creating database with default mode of this attribute. New in TRS 4.00(BUILD:1011): 1. Optimize the AND class retrieval operation to cut down opera- ting load. 2. Limit the max length of record data to avoid that the OS will be busy with swapping virtual memory when uploading database. 3. Support inserting a BASE element automatically into HTML file when uploading database records. 4. Support cheching of reserved words to avoid using them as TRS object name. 5. Support optimizing databases automatically when TRS Server is idle (no work to do). 6. Support comments in some TRS format record files. New in TRS 4.00(BUILD:1010): 1. Resolve the security problem that a illegal user can destroy the share LOGIN table. 2. Enable the record crypt of database without public selecting attribute, if the user is not the super user, or a DBA, or a GDA, or the owner of database. 3. Process local connection (e.g: shutdown) with high priority. 4. Support Altering data file path of database. New in TRS 4.00(BUILD:1009): 1. Support option of killing Server when shutdown. 2. Resolve the overflow problem of user instances made by login of web applications, and the instances are never logged out. 3. Resolve the problem that makes Server crash (system resource deadlock, can't be shutdown) on some UNIX platform since too many child processes forked (exceeding the maximum number of possible open files for per process).