THE SET ANALYSIS Summary 1 Why use the sets. The identifier. ‘The operators... Ron The modifiers. 4.1 All members ... 4.2 Known members 43 Search string 4.4 Using a boundary for an integer dimension 4.5 Using a variable... 4.5.1 Using a variable storing one or several members... 4.8.2 Using a variable storing a search string... 45.3 Using a variable storing integer members... 4.8.4 — Using a variable storing a dimension... 4.8.5 Using a variable storing the whole set... 4,6 Using a function... 4.6.1 Numeric function .sssnnsssenese see 4.6.2 Function retuming members... 4.7 Using p() and e() 4.8 Using two fields... 4,9 Using IF statement... 4.10 Mixing Alternate States.... 4.11 The extreme results... 4.14.1 The whole application... sees 4.11.2 The empty sets... 5 Other links... Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 1Foreword ‘This documents also availabe in French even if the French one has yet the version 1.0. btpscommunty alivien.con/dors/DOC-488 ‘Tho 4 paragragh answers fo one spec question: How do you wie the sol analysis whon you have a variable, a search boundaies or? feds ving, whon you have doc: area book woul! be ‘Tho aim to clarity some syntax ofthe Sot Arabyss, iis not a comp 1d lo cover the Sot Analysis part of ‘ieview. But nope vl be usetul Enjoy The version of this document is 2.1. It was last modified on 25FEB2014. QlikView 11.20 SR2 has been used, Short naming convention in this doc: YOXOK_KEY: dimension key XXX_LDESC: simersion ong descrinton XXX_SDESC: mension shor ceserpton The TIME axis contains: ~ Year, Month - _TIME_KEY, TIME_SNAME et TIME_NAME that are periods from 2010 to 2013, current month. A concatenation of years and months. TIME is a sequential (linear) dimension while Year and Month are cyclic ones. Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 21 Why use the sets The sets let you create a different selection than the active one being used into the chart or table, The created group let you compare the aggregations of this group and the one of the current selection. These different aggregations let you compare for example the sales of the current month and those of the. previous year, create YTD or moving totals. You may also create market shares, a percentage of the Products, of the year A set modifies the context only during the expression that uses it. Another expression without any set will get the default context, standard selection or the group of the Alternate State. Asset is always written between curly braces {...}. A set can modify the selection of one or several dimensions. It can be composed of a single set but can include also an inner set. Aset may be composed of - Anidentifier = An operator - Amoditier These 3 fields are optional. Of course, to write a set, you will need to incorporate one or several of these fields, le Limitation: take care, a set analysis Is calculated once per chart (or table). NOT once per row. If you want a different result in your set analysis according to the row being calculated and displayed, it will NOT work, Change the way to compute your data (model, if statement) but forget the set analysis Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 32 The identifier 0 - empty set 1—complete application $- current selection (take care that sometimes your chart or table does not use the current selection but an Allernate States) $1 — previous selection ($2: the second previous selection, etc.) $_1—next selection Bookmark01 — name or ID of a bookmark Group — group name (Alternate State) ‘Sum({Book1} [Sales}) : sales of the selection of Book? Sum({Group1} [Sales}): sales of the Alternate State Group1 (the syntax is identical) Sum({1} [Sales}) : sum of everything (All dimensions are completly reset to All) ‘Sum({$} [Sales]) : sum of the current selection (= sum({Sales]) 3 The operators + :union of sets * Intersection of sets ~ : Exclusion of two sets 1: members belonging to only one of the two sets Ifyou use the operators with the equal sign: : add the set to the current selection : remove the set from the current selection ‘Sum({1-$} [Sales]) : sum of the sales of the « database » except the current selection ‘Sum({GROUP4 * Book1} [Sales}) : sum if the sales for the intersection of Group and Book1 (the members belonging to both) Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 4The sets that we will study are between angle brackets : <>. They are quite more complex also. But you may use also the operators: { -} or {set 1}-{set 2}: members in the set! that are not in the set2 (because we remove them) 4 The modifiers Yes, that is the tricky part of the Sets Analysis ! With them, you can reduce or enlarge the scope of the selection used in an aggregation. You may modify the selection for one or several dimensions. A set analysis is not like clicking in the ListBoxes. If you decide to set a dimension to a specific value currently not selected: = Inthe ListBoxes, you will get a new selection - Inthe set analysis, you will NOT: you will just set this dimension to this member (and the intersection with the current selection may be empty) A set analysis is calculated before the chart is computed and redrawn. As you may natice, itis NOT calculated per each row/column in the chart. So, you may not guess a scope of product depending on the rows that are products. You may not guess another set of periods depending of the columns if the Period dimension is in column : the YTD or moving totals you may encounter are often valid only if the period is not a dimension of the chart 4.1. All members selection } ({*} for numeric, ("*"} for text) Ling selection ) will get the year 2012 but only for the months that have been selected in the ListBox (that is not the Total of the Year 2012) {} will get ALL months of the year 2012 Each time you have a hierarchical dimension and display several Listboxes, take care to reset to all every dimension except the one you want to select. By setting a value of one dimension, you do not select any associated members (as QlikView does naturally with the ListBoxes).. 4.2 Known members Syntax Selection Bx: {AMBOISE, BELLE, AUTREMENT}>} {AMBOISE, BELLE, AUTREMENT), CATECOR\ In the global syntax: sur (() [Valse Please note that The selection may be the current one (S sigh), the database (1), a bookmark or a group. The current selection is the default selection, the $ may be omitted. = There is no comma (,) between the set and the measure name - The dimension or the measure name must be enclosed into brackets (| ifthe name contains specific characters like a space, a hyphen (-) - Members are always enclosed into curly brackets (} : whatever the way to find them (by name, by search string, by a function) - Named members are separated with a comma (,) - Numeric member names are not enclosed into single or double quotes - Text members are enclosed into single or double quotes but not necessarily if they do not contain specific characters like spaces - is empty: ye} If we compute some flags (1 or 0 according to a test) into the script, we may use them easily in the set analysis: Sum(i (} ‘And because there is no specific character in this member name, | may write ita 4th way: {) This technique is very useful when we include a function including a set into a more global set. 4.3 Search string Sometimes, we want to get all MANUFACTURERS with a specific name or which contain specific letters in their name: Syntax: () + 1 =0to several characters (e1+ will etum P1, P10, P11...) * = character (217 will return P10, P11 ... but not P1) We can do several search strings in the same set: mtaxe : { - or as explained earlier by placing nothing at the right of the equal sign . The two following sets are identical if you do not have any NULL value in your field. you have any NULL value in your field and want to select it also, choose . In the database used as example, there is no NULL dimension value: ONE category 1 CATEGORY_LDESC=("ACC") >} {} Ex: ALL MANUFACTURERS except those whose name bs [ whose name beg Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 7Take care: () and ()_ are very different, The first set removes from the current selection all MANUFACTURERS whose name begins with ENT, while the second set selects all MANUFACTURERS except those who begin ENT : the {°" is by default x: ALL MANUFACTURERS, Alle categories whatever the selections {) : ALL years (because it is numeric, nc quote), all Fabricant = Manufacturer mension of the tableau. Category Ls not Ginension: it has been set Tableau par fabricant (Ventes Volume) FABRICANT [Name @ Somme delase.,. Total categcries — % Categerie ‘AUTRES FAERICANTS 9z7 tzi 1977 973 46.87% 53,48% | TTS 10820 71546 15,26% 0,83% Na 361271 1.498 685 27,10% 21,99% 409735 425 648 96, 04% 23.69%, 141 141 100,00% 0,01% | 663, 663 100,00% 3.04% ( ( 4.4. Using a boundary for an integer dimension Most of the time, Years are numeric. That is also the case for many keys because QlikView (as SQL) is faster when joining numeric keys than string keys. The syntax is quite strange because you need to insert the code between double quotes: (}: Allthe TIME_KEY members less than 12 (=12 <=18")>): key members between 12 and 18, both included x: keys 4, 5, 6 and those b ( Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 84.5 Using a variable 4.5.1 Using a variable storing one or several members (8 (vamevariable)} > sign in front of the variable) Please note that $/=NameVariable) will interpret the variable and therefore QlikView rewrites the command before executing it. The content of the variable may contain several members, as if they were in the real syntax. Enclose them between quotes or bracket according to their name. Setings for Seectec Vaeble Value AMBROSIENTREMONT —() Ex © there is another $ signin the set analysis that means the current selection ¢ (ameVarsabie) it ean be enclosed between double quotes or not lke in the example above. If there is no quote, you may add some in the text of the variable because the member names contain spaces. Setingsfor Selected Varable Value "AUTRES FABRICANTS:, Al] 48.2 Using a variable storing a search string Anormal set: () The variable vChoiceProduit stores a search string and also the double quotes “E™ () The periods between two dates: (er mm KEY = { $(vFirstPeriod) <¢(=vLastPeriod)"} 4.5.4 Using a variable storing a dimension We have already seen that we can use variables with the § sign. This variable may contain members but also a dimension: Ex: sum( {S <$(vDim)-i"*"} >) [Volume Sales]) Because the dimension may contain spaces or specific characters, it would be safer to use the square brackets (remember that QV rewrites the syntax, so once your dimension name with a space replace the variable, you would add brackets): sun (5 <[$(woim)]=("™") > es] 4.5.5 Using a variable storing the whole set ‘We can store members or dimension into a variable, Why not put the whole syntax? Ex: sum({$ <$(wSet) >} [Volume Sales]) Ex: sum({$ <$ (Set), CATECORY_LDESC= As you see, we can store the complete syntax or just part of it! " >) [Ventes Volune)) The vSet variable must contain a valid syntax except <> like URER, ), TIE si 01/33") + Year={$(-max((1} Year))), TIME LD! 4.6 Using a function So far, we have selected members that we knew in advance they will be or not in the set. But we can also Use funetions that will return members according to a test, a rank 4.6.1 Numeric function Syntax: Please note’ ~The max function, like any aggregation function, take into account the current selection - We have used an inner set. {1} inside the max function means the whole application. If we do not Use it, the max function will return the last year among those chosen by the user - We use two equal signs, 1 before the curly braces, 1 between the quotes Ifyou are looking for the previous year of the last year of the selection, you may write one of the two syntaxes that are equivalent: = , (S}= current selection, it is the default Most of the time, the set analysis become tedious, especially when writing YTD, Moving Total, Comparisons vs Year ago. One way to simplify the syntax is to populate some fields directly into the model, Herebelow, the field A-1 stores the key for the Yr Ago period, P-1 stores the key for the previous period, YTD stores a flag (1 or 0) to indicate if the period should be taken to compute the YTD data: TINE KEY Mon ven At pa yD 12. denver na L 2 1 15 Fever 23 2 3 1 15 mars 2s 3 4 1 16 Avr nB 4 5 1 a7 Msi ne 5 6 1 18 lun ne ‘ 7 1 19 ullet ne 7 8 ° 20 At 2s 6 3 ° 21 septembre 2s 8 20 ° 22 octobre 2s 0 2 0 23 Novembre 2013 1 2 ° 24 Décenre ne 2 2 o 25 Janvier nyo 24 1 26 Fei nyo 25 1 27 Mers nw 5 6 1 you want to get the periods of the Year Ago, you just need to use the appropriate field and the concat function: sum ((} VALEUR) YTD: sum({) VALEUR) YTD from Year Ago: sum({) VALEUR) MOVING TOTAL 12 periods: sum({S(+Max(TIME_KEY)-24) <=$(+Max(TIME_KEY}'2)">) VALEUR) Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 11é 4 Keep in mind that the set analysis is “computed” once before the chart is computed. So, if you are looking fora set that returns differnt members according tothe row or column of the chart. you may watt fora long while Therefore, you can also modify deeper the model so that the set analysis for all the YTD, Moving total, Year ‘Ago comparisons become very simple, See a doc | have written about this possibility: http:icommunity.qlikview,com/does/DOC-4821 Some people populate flags (with 1 and 0) so that they can write : Sum(VALUE*Flag) This is simple and works well. But itis very time consuming, much more than sum({ VALUE) because QlikView in the second case reduces the scope before the computation. With sum(VALUE*Flag), likView does not scope the data and computes everything. See the excellent answer by John Witherspoon in the following discussion: ttp://community.glikview,com/messaqe/1434 4.8.2 Function returning members The text function is between quotes. No $ sign. ntax: ‘We want to aggregate the volume sales for the MANUFACTURERS whose value sales is greater than 100,000 dollars based on the current selection: sum{ [S 100000"} >} [Volume Sales!) But we can also decide to search the MANUFACTURERS on a specific period, or for specific products. We will create an inner set: sum( (3 ) [Volume Sales}}>100 0") >} [Volume Sazes}) Remember that the members could be enclosed between single or double quotes. Because the function uses double quotes, we will use for the members either the single quotes or the square brackets (]. [want to find the Manufacturers whose sales are over 100,000 for the 2 categories CHEESE CAKE and ACC in January 2013 (period P01/13). But I want to remove from that list those whose sales are lower than 50,000 for all categories in January 2012: Ineed two sets : { - }, each of them will use the function sum(): Set ("sum (11< Value Sales))>50 *p 01/12"), CATSGORY_LDRS: ter Ina global syntax 2 ) [Value Sales})>S0000") > mesum({1¢ TIME sp 01/12"), cm Value this could be reduced a little : sum( (1 ) [V Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 1201/12"), CATEGORY_LDESC=(**'})>) [Value Sales]) <500) >} es] Please note that the function is between quotes: you will have no help to write the syntax, for the function, nor for the fields. QlikViow is case sensitive for the field names: i you write TimeKey instead of TimeKEY, ‘you will get 0, So test the function before including it in the set analysis. Ityou want t get the Top 20 products: Sum ({} sales And now the Top 20 products of the brand 10 (we need to use an inner set): Sum({} sal £ ‘Take care NOT to write: m({} sales) ‘You would get the intersection of Brand 10 and the top 20 products. And perhaps, you will get nothing because there is no product of the brand 10 among the top 20 products. © oop also in mind thatthe set analysis is “computed” once, before the chart is computed. Not once per row: you will not get different products by region if you have a region dimension in your chart 4.7 Using p() and e() These functions retums the members that have (or have not) data. They are based on the associative model. al }>) [, Other Syntax: {} Please note that: - Both functions p() et a() have the same syntax : one p() returns the possible values, the other e() retums the excluded values - They will be inserted into a greater set ‘These functions do let you search the Top N products or those whose sales are greater than X : they just return the members (ie the products) that have data, or those that have not. Ex: we are looking for the manufacturers who sell CROISSANT in the first period of 2013. But we want to display all categories: Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 13pi) can also be used when you have two dimensions, with the same values, that are not linked. And you want that a selection made on first dimension is also propagated to the second one. For example, | have 2 dimensions ClientID and BClientID that have the same values, but are not linked The user has selected the ClientID through a ListBox and we need to display the Budget related to BClientiD. sum ( 4.8 Using two fields If we want to select a dimension through two fields, for example: ~ Billed client = Delivered Client - Ordering Date = Delivery Date The syntax is more or less the one seen in 4.6.2. section: The boolean condition will be created by comparing the two fields. © © stontion: the searched dimension cannot be also in the boolean condition. If needed, create an integer key with Autonumber(). ‘We want to get the sales that have been delivered tne same day. We have two fields : DayDelivery and DayOrder Sum ({) Sales) If the user chooses several products: Product P Set Analysie dana variable avec IF L Set analysis ax.o Product © Sabs 2 80.2 5 3 121 6 4 8.8 But if the user chooses one single produ Product P — setanalysis dans varietle aves IF 1 2 set analysis, ano product sacs 4 1 1372 E 2 1902 B 5 1 4 1258 5 136.6 é Ey Remember that the set analysis is computed once per chart, so it is useless to get a IF statement that would return a different scope across the cells (rows, columns) of the chart. In other words, the IF statement cannot return a different value according to the tow displayed in the chart. Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 154.10 Mixing Alternate States An Alternate State is a way of creating several groups of data to compare the two as a group. For example, you may want to compare two groups of customers (or products ....): these groups will be chosen by the user through a ListBox pointing on a unique field Company 2 Company 2 pe 5 c c 5 D E £ As you can see, we get 2 different selections for the Company field. They are not linked any more. These two fields may be displayed into 2 several sheets if you want to offer ni different sheets. linked selections in the How to create the Alternate States: 41) Settings / Document properties / button Alternate States (and then create the different groups you may want to offer the user) 2) For an object: edit the object / choose the Altemate State te Atemate Siste Object ID Gla = 106 Take care, this Alternate State combo box is not displayed if you did not perform the first step. 3) For the sheet, same way to do. But for the other objects of the sheet, if you want them to get this Alternate State, choose . It is an easy way to change the State in just one place. Using a single Alternate State Syntax: { (NameO£TheAlternate: te] ), Sales) Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 16Using several Alternate States If you want to compare several groups of companies, you will display at least 2 ListBoxes for Company using several Alterate States. But, what for the other dimensions like Year, Month, Products ... Will you need to display them also in double? The answer is NO, but once again, the syntax is quite strange: {{Name of the Alternate State] } Ex. (3 <{Year] = (Als: (vear)>) Here above, I want to get the Alternate State B, except on Year where I want to apply the Alternate State A. Please note: - Ifyou want to compare to the standard selection, you will the $ (that is in fact a specific state of the selection - You enclose the Alternate State between [] only if the name contains specific characters The dimension after the two semi-colons (:) is also between brackets [J or nothing depending to it's name (here, we could have written - You need to repeat this syntax for each dimension you want to share with the other Alternate State (Month, Week, Day, Product ...) With this syntax, you should (or perhaps only could) avoid to copy the selection of A into B for several dimensions. In a graph, color in red data of the companies that belong to (A} Alternate State. The other companies will be in blue: iftmateh( |" & Sy red, blue) } al Saes Eackatoune: Color PeePopeut For a more complete brief on Alternate State, please review the doc: http://community.alikview.com/docs/DOC-3837 Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 174.11 The extreme results 4.11.1 The whole application You can use * between the braces (}: {1} sum((1} [Sales Value]) Last year (we already studied it): $ (Max ({1} YEAR) ) To get everything, except one or two fields, see the excellent post by Smakovskiy bhito:/community.alikview.com/docs/DOC-1334 If Lwant everything (All values) except for 2 dimensions: PRODUCT and MONTH, | would write: =sum(S-<{8(=Cone ((1<$Field-=(PRODUCT MONTH > distinct $Fleld: Js, =} VALUE) By using the $Field in the concat function, Qlikview writes: [dim1}=, [dim2]=, [dim3]= .. resetting to all values all the dimensions except the one inside the curly brackets {), Did you notice the Field -= {) that tolls QV to take everything except the content of the bracket? Assimpler way to write: sum({1} Budge!) OR sum((t) Budget) To get al selections except on Region (current selection) 4.11.2 The empty sets It can happen that the set is empty. The result will be 0. Its either something wanted, or something more unwanted - You use the braces without anything in it: {) - You typed the fields or the members incorrectly: Qlikview is case sensitive - You typed a wrong expression - You typed a wrong intersection: a product, a client but this client does not buy the product (0 is therefore the right answer) - You forgot to reset a field that is limited through a ListBox. For example, you have brands and products in Listbox, and the user select a product. By setting a brand in the set Analysis, you may Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 18encounter an emptiness because the chosen product does not belong to the brand. You need to reset to all the producls: {} For time periods, when we want a total year, it would safer to write in the set analysis all the dimensions that could be set through ListBoxes: um({$ [Volune]) This time, we get the whole year, whatever the selection the user has made. Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 195 Other links ‘A GUI that generates some code (a good way to learn): http:/tools..likblog.at/SetAnalysisWizard/QlikView-SetAnalysis Wizard and Generator aspx?sa= Thad the pleasure to write some other documents that may interest you: LOAD data into QlikView htto://community.alikview,com/docs/DOC-5698, Set Analysis, French version of this document: httod/community alikxview,com/docs/DOC-4 ‘Simply create YTD, moving totals and comparison vs Year ago: htip:/community .likview.com/docs/DOC-4824 Dynamic tables and graphs: In English: http:1/community.alikview.com/docs/DOC-5733 In French: http://community ali«view.com/docs/DOC-5735 Fabrice AUNEZ — version 2.1 The Sets Analysis-Page 20