Lotuscript for large systems
0 likes828 views
This document provides tips for writing LotusScript code for large systems with a focus on logging, performance, code reuse, and handling weird situations. Some key points include: - Logging is important for stability and managing large systems. Recommends using OpenLog or creating and emailing log documents to avoid performance impacts. - Views with click-sorted columns and unnecessary views hurt performance. Recommends minimizing views and avoiding click-sort. - Agents need to be well-behaved to avoid overloading servers. Suggests profiling agents, breaking large tasks into multiple runs, and not relying on Agent Manager to kill misbehaving agents. - Code reuse is important for maintenance. Recommends creating
![Weirdness: Execute
• You all know how to do ‘Evaluate’ right?
Dim V as variant
V = evaluate( “ @Name( [CN]; @UserName ) “ )
print V(0)
‘ Result is ‘Bill Buchan’](https://image.slidesharecdn.com/lotuscriptforlargesystems-130205140628-phpapp01/85/Lotuscript-for-large-systems-52-320.jpg)
Lotuscript for large systems
- 1. LotusScript Survival Tips for Large Systems Bill Buchan hadsl.com
- 2. Agenda • Introduction • What do we need? • Logging • Performance • Code Reuse • Weirdness
- 3. Introduction • I am: • Bill Buchan, CEO, hadsl • LotusScript expert since 1997, Large systems experience
- 4. Direction? • Lets consider the following slide...
- 5. Whats New In LotusScript? This page is left intentionally blank
- 6. Whats new in LotusScript? • The Class libraries have generally grown with each release • Leveraging feature-specific stuff such as ID Vault in 8.5.0
- 7. What does this mean? • LotusScript is still an invaluable tool in your armoury for creating Business Logic • Its easy, its multi-platform • Its NOT the future in terms of Client or Web • Check out all the xPages sessions here • No excuses
- 8. What do we need? • Large environments need: • Stability • Scalability • If you cant measure it, you cant manage it • You want to do more, with less time and less people
- 9. What do we need? • Stable performance • We cant keep adding new servers every time we add a new application/database instance • We cant let a single application wreak havoc on our systems
- 10. What do we need? • Code Reuse. Of course. • We have less resources doing more work • We need to test more effectively
- 11. Logging • Why do logging? • As a rule of thumb, if its scheduled, critical, or running in someplace you cant immediately check, you need logging • You need to know: • When and where did it start, stop • Was it successful
- 12. Logging • If something happens, and you find out from the users • You have failed
- 13. Logging • Lots of solutions available • For instance, use ‘msgbox’ to log to the console. • Admins hate this • It’ll never be acted upon
- 14. Logging Solutions • Check out OpenLog at OpenNTF • Add two lines of code, and you have a comprehensive log • Free and Easy • No excuses
- 15. SuperLogging • Consider Performance • Opening a notes database - even from Script - opens the default view • Log database views can get large • It can be slow
- 16. SuperLogging • Why not • Create a new document and populate as normal • Stamp it with start and stop times - handy for measuring load • Add Server, Database, Agent and User • But then use notesdocument.send()
- 17. SuperLogging • Sending log documents by email means: • The process is not waiting for you to save the log document (and update the views) - huge performance increase • The log database need not be local • The user processes do NOT need access to the log database
- 18. SuperLogging • Once you start emailing in Log Documents, you can do more: • Spot when agents DONT trigger a log document - and do something about it • Spot when NEW agents start creating logs
- 19. SuperLogging • Remember, ALL code has to log accurately • LotusScript, Java, xPages • Javascript?
- 20. Performance • We application developers can create fantastic applications • And we can create horrible ones • Ones that really kill servers. • Lets talk about...
- 21. Performance:Views • Your application needs views. Of course it does. But does it need them all? • ‘Click-sort’ columns are evil • Use more ‘Categorised’ views
- 22. Performance:Views • How can I tell the size of a view? • Admin client, select database, Tools, Manage Views
- 25. Performance:Views • Create as few views as possible • Don’t use Click-sort on columns unless absolutely necessary. Each click-sort column creates another copy of the view
- 26. Performance:Agents • Scheduled Agents have to be well-behaved • Agent Manager doesn't really do a great job in killing runaway agents • It certainly doesn't throttle agents which consume lots of CPU • And it only runs one or two agents at a time
- 27. Performance:Agents • So how to we minimise our Agent Footprint? • Write them to fit the potential size of your application • If there are lots of documents to process, let the agent run multiple times
- 28. Performance:Agents • Use Agent Profiling to find out how long your agents take against a representative quantity of documents • ‘It worked in testing’ is not an excuse. • Its an admission your testing wasn’t rigorous enough
- 29. Performance:Agents • To enable Agent Profiling, check the ‘Profile This agent’ in Agent Properties. • Once its ran, you can view which calls its made, and how much time it took
- 31. Performance:Agents • You can see • When the agent ran • Where it ran • Where it spent time • Invaluable for performance tuning
- 32. Performance:Agents • Now. Lets take it one step further • How about collecting ALL agent profiles from ALL databases and storing them in a database? • NOT recommended for production environments
- 33. Performance:Agents • Recipe: • Take ‘TriggerHappy’ from OpenNTF • Have it listen to ALL document saves on the server • If the form is called "$BEProfileR7" • Forward the document to another database
- 34. Performance:Agents • For more information on this, visit: • http://www.billbuchan.com/ imported-20091119232548/2010/2/2/ universal-agent-profiling-in-domino.html • URL is in IBM Standard Download Notation.
- 35. Performance:Agents • A common scenario • We build an agent to examine a lot of documents, and update a few of them, every night. • The database grows • Your admin kills you.
- 36. Performance:Agents • Remember: • Agent Manager is like a crumbly old relative - Everyone loves him • But he’s quite shaky on his feet • And smells • Don’t abuse him
- 37. Performance:Agents • Changing the maximum agent run-time on the server document is CHEATING
- 38. Performance:Agents • Re-architect the agent • It starts up, figures out how long it has to run, and works out its termination time. • Maximum time minus 2 mins • Keep working till expiry time • Have the agent run a few times - say every 10 mins for 3 hours overnight
- 39. Performance:Agents • So: • Own & Understand your agent performance • Don’t abuse agent manager • Don’t let it kill your system
- 40. Code Re-use • Don't reinvent the wheel every time • That leaves far too little time for the pub • Oh - and it means re-testing from scratch every time • Be Lazy
- 41. Code Re-use • So • If your an old-fashioned script guy, create a standard script library that encapsulates: • Standard Logging • Configuration lookup
- 42. Code Re-Use • If your a little bored with this (and who isn't?) • Create a LotusScript class which encapsulates all boring logic • Extend this class in your own code for the business logic • Simplify the target code
- 43. Code Re-use • For example: class baseClass Public Sub logMsg(msg as String) ‘ do stuff end sub public sub raiseError() ‘ do stuff end sub end class
- 44. Code Re-use class myLogicClass as baseclass Private Function myStuff(waffle as String) on error goto errorhandler call logMsg(“hello world”) ‘ Do stuff exitFunction: exit function errorhandler: call me.raiseError() resume exitFunction end function end class
- 45. Code Re-use • Remove the responsibility for • Logging • Error handling • Other boring stuff • From your business logic
- 46. Weirdness • Sometimes, we have to walk on the wild side • Just sometimes, we like to stroke our misshapen little ego • Or we actually have a business reason • Whatever
- 47. Weirdness: Lists • Use lists - they’re not hard. Dim Pres list as String Pres(“George”) = “Bush” Pres(“Bill”) = “Clinton” Print “Georges’s last name is: “ + Pres(“George” ) if not isElement(Pres(“Chad”)) then _ print “Chad wasn’t found” forall thisPres in Pres Print listtag(thisPres) + “ “ + thisPres end forall
- 48. Weirdness:C-API • LotusScript never does everything. Sometimes you have to call out to C-API • What is C-API? • Over 700 functions that you can get to using C code • Downside: Creating & Supporting code on every platform you support. Eugh.
- 49. Weirdness: C-API • So how can the lazy programmer call C-API functions from LotusScript? Declare Function W32_NSFGetServerLatency Lib LIB_W32 Alias {NSFGetServerLatency} (_ Byval ServerName As Lmbcs String, _ Byval Timeout As Long, _ retClientToServerMS As Long, _ retServerToClientMS As Long, _ ServerVersion As Integer) As Integer
- 50. Weirdness: C-API • Now lets exploit this declaration Public Function getServerLatency _ (strServer As String) As Long Dim nnServer As New NotesName(strServer) Dim ToServer As Long, fromServer As Long Dim ver As Integer Dim timeout As Long timeout = 1000 1000ms == 1 second ' Call W32_NSFGetServerLatency(nnServer.Can onical,_ timeout, _ toServer, fromServer, ver) ' Return both directional latencies added togeth er getServerLatency = fromServer + ToServer End Function
- 51. Weirdness: C-API • Now do this for every function, and every platform you wish to support. • Euch
- 52. Weirdness: Execute • You all know how to do ‘Evaluate’ right? Dim V as variant V = evaluate( “ @Name( [CN]; @UserName ) “ ) print V(0) ‘ Result is ‘Bill Buchan’
- 53. Weirdness: Execute • Boring. Lets write some code to write some code Dim S as String S = “ msgbox | Hello World | “ V = execute ( S )
- 54. Weirdness: Execute • What is this doing? • We can populate a string with valid Lotusscript Code • And then execute it. • Why on earth would this be useful? • We could, for instance, generate our platform specific C-API calls
- 55. Weirdness: C-API • Check out the ls2capi.com web site • And check out my ls2capi database on • http://hadsl.com/hadsl.nsf/Documents/ LS2CAPI+-+Calling+Notes+C-API+from +LotusScript!OpenDocument • Step through it in the debugger.. ;-)
- 56. Conclusion • Use these techniques to deal with large systems • Minimise, Simplify and Productionalise your code • Learn xPages