DR 16-0007 - SML: omissions and inconsistencies in the specification of attributes (Part 2)
Francis Cave
francis at franciscave.com
Sat Aug 19 10:42:52 CEST 2017
Murata-san
I agree that changing attributes from optional to required is risky, so we should certainly consider alternatives to my proposals. However, I believe that in all such cases the omission of the attribute renders the element meaningless, in which case we have to decide whether allowing conformant documents to contain such junk is acceptable.
Kind regards,
Francis
Sent from my iPad
> On 19 Aug 2017, at 04:53, MURATA Makoto <eb2m-mrt at asahi-net.or.jp> wrote:
>
> Francis,
>
> Thank you very much for your thorough work.
>
> Some existing optional attributes are made mandatory in your
> proposal. We have to make sure that this change does not
> cause any problems to existing documents.
>
> I created a text file, available at http://www.asahi-net.or.jp/~eb2m-mrt/AttsPerElem.txt
> It provides a list of always-present attributes and that
> of sometimes-present attributes for each element. This
> file is created by examining all OOXML files in my computer.
>
> I hoped that my experiment would help, but none of the elements
> having optional-but-proposed-as-mandatory attributes appear in
> my file ......
>
> Regards,
> Makoto
>
> 2017-08-18 1:04 GMT+09:00 Francis Cave <francis at franciscave.com>:
>> Dear all
>>
>>
>>
>> Here are proposed resolutions and actions for a further 12 elements in DR 16-0007.
>>
>>
>>
>> Francis
>>
>>
>>
>>
>>
>> §18.8.7 cellStyle (@name) (complex type: CT_CellStyle)
>>
>>
>>
>> There is a note on this attribute in MS-OE376 (limiting the name length – see section 2.1.724, page 310, note a), but no mention of the attribute being optional in the standard or how to interpret its omission.
>>
>>
>>
>> I pointed out on 2016-08-08 that this issue relates to DR 16-0016. Any further clarification of the prose will need to take account of the resolution of DR 16-0016.
>>
>>
>>
>> On further reading of the current specification, and after running some tests in Excel, I conclude that @name is optional because it is not required if @builtinId is specified. . See Annex G.2 for a list of built-in style names. If @builtinId is specified and @name is either omitted or has a value other than the one specified in Annex G.2, Excel opens the document without complaint and restores the correct built-in style name when the document is saved. If both @builtinId and @name are omitted, Excel opens the document without complaint and silently discards the unnamed style.
>>
>>
>>
>> Proposed resolution: Add the following to the description of @name in prose: “If this attribute and builtinId are both omitted, this cell style shall be ignored.”
>>
>>
>>
>> §18.8.34 rgbColor (@rgb) (complex type: CT_RgbColor)
>>
>>
>>
>> There is no note on this attribute in MS-OE376.
>>
>>
>>
>> The element rgbColor has only this one attribute and no content, so is meaningless if the attribute is omitted. Yet the schema explicitly specifies that the attribute is optional. The complex type CT_RgbColor is only used for this element. I therefore conclude that there is a mistake in the schema.
>>
>>
>>
>> Proposed resolution: Change the schema to make this attribute required. No change to prose.
>>
>>
>>
>> §18.10.1.2 b (@cp, @f, @u) (complex type: CT_Boolean)
>>
>>
>>
>> MS-OE376, subsection 2.1.763 (page 339) includes the following notes:
>>
>>
>>
>> c. The standard does not specify a default value for the u attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> d. The standard does not specify a default value for the f attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> …
>>
>>
>>
>> g. The standard does not define the relation between the u attribute and the sharedItems parent.
>>
>>
>>
>> In Office, the u attribute only applies when the parent is sharedItems.
>>
>>
>>
>> h. The standard does not define restrictions on when the cp attribute should be present.
>>
>>
>>
>> In Office, the cp attribute applies to OLAP-based PivotTables only and the parent is sharedItems.
>>
>>
>>
>> i. The standard does not specify restrictions on value of the cp attribute.
>>
>>
>>
>> In Office, the value of the cp attribute shall be equal to the number of x child elements and less than or equal to the mappingCount attribute of the corresponding cacheField element.
>>
>>
>>
>> j. The standard does not define the relation between the f attribute and the sharedItems parent.
>>
>>
>>
>> In Office, the f attribute only applies when the parent is sharedItems.
>>
>>
>>
>> The attributes @f and @u indicate binary properties of an item in a PivotTable: is a calculated item or not; is an unused item or not. Omitting the attribute would logically imply that these properties don’t apply to the item, so I conclude that Excel’s usage is the only logical usage and should be documented in the standard.
>>
>>
>>
>> Should some of the Excel restrictions on the use of @cp, @f and @u be documented in the standard, or are they implementation-specific?
>>
>>
>>
>> The complex type CT_Boolean in SpreadsheetML is only used for this element, but this element appears in content models defined by the following complex types:
>>
>>
>>
>> CT_GroupItems
>> CT_Record
>> CT_SharedItems
>>
>>
>> MS-OE376 states that @cp is only used by Office in sharedItems (§18.10.1.90). Strangely, it is only groupItems (§18.10.1.38) that explicitly mentions OLAP. There are no examples in the standard of the use of @cp, although it is available on six SpreadsheetML elements: b, d, e, m, n and s. Judging by the notes on these elements in MS-OE376, the usage of @cp varies according to the element, so examples in each case would be useful to promote interoperability.
>>
>>
>>
>> Proposed resolution: Add the default value false to the schema for both @f and @u. Provide an example of the use of @cp on at least one of the six elements on which it is available.
>>
>>
>>
>> §18.10.1.8 calculatedItem (@formula) (complex type: CT_CalculatedItem)
>>
>>
>>
>> There is a note on this attribute in MS-OE376 (limiting the maximum value – see section 2.1.769, page 343, note a), but no mention of the attribute being optional in the standard or how to interpret its omission.
>>
>>
>>
>> This element specifies a formula associated with a specific PivotTable field. It does not make any sense for the PivotTable field not to be specified. However, the presence of extLst in the complex type CT_CalculatedItem suggests that there might be other ways in which associated the PivotTable field could be determined.
>>
>>
>>
>> Proposed action: Ask the Excel team to clarify under what circumstances @field can be omitted. Depending upon their response, decide whether or not to change prose or change the schema to make this attribute required.
>>
>>
>>
>> §18.10.1.21 d (@cp, @f, @u) (complex type: CT_DateTime)
>>
>>
>>
>> MS-OE376, subsection 2.1.777 (page 346) includes the following notes:
>>
>>
>>
>> a. The standard does not specify a default value for the u attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> b. The standard does not specify a default value for the f attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> …
>>
>>
>>
>> g. The standard does not define the relationship between the cp attribute and the datasource type.
>>
>>
>>
>> In Office, the cp attribute applies to OLAP-based PivotTables only.
>>
>>
>>
>> h. The standard does not define the relation between the cp attribute and the sharedItems parent.
>>
>>
>>
>> In Office, the cp attribute applies when the parent is sharedItems.
>>
>>
>>
>> i. The standard does not define the relation between @f and the sharedItems parent.
>>
>>
>>
>> In Office, @f applies when the parent is sharedItems.
>>
>>
>>
>> j. The standard does not define restrictions on the value of @cp.
>>
>>
>>
>> In Excel, @cp shall be equal to the count of child x elements and be less than or equal to the @mappingCount of the corresponding cacheField element.
>>
>>
>>
>> k. The standard does not define the relation between @u and the sharedItems parent.
>>
>>
>>
>> In Excel, @u applies when the parent is sharedItems.
>>
>>
>>
>> See discussion above for §18.10.1.2 b.
>>
>>
>>
>> Proposed resolution: Add the default value false to the schema for both @f and @u. Provide an example of the use of @cp on at least one of the six elements on which it is available.
>>
>>
>>
>> §18.10.1.27 e (@bc, @cp, @f, @fc, @in, @u) (complex type: CT_Error)
>>
>>
>>
>> MS-OE376, subsection 2.1.781 (page 348) includes the following notes:
>>
>>
>>
>> d. The standard does not define the relation between the cp attribute and child elements.
>>
>>
>>
>> In Office, the value of the cp attribute shall be equal to the count of child elements and be less than or equal to the @mappingCount of the corresponding cacheField element.
>>
>>
>>
>> e. The standard does not define the relation between the cp attribute and the datasource type.
>>
>>
>>
>> In Office, the cp attribute applies only to OLAP-based PivotTables.
>>
>>
>>
>> f. The standard does not define relation between the cp attribute and sharedItems node.
>>
>>
>>
>> In Office, the cp attribute applies only when the node parent is sharedItems.
>>
>>
>>
>> g. The standard does not specify a default value for the f attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> …
>>
>>
>>
>> i. The standard does not specify a default value for the u attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> …
>>
>>
>>
>> k. The standard does not define relation between the f attribute and sharedItems node.
>>
>>
>>
>> In Office, the f attribute applies only when the node parent is sharedItems.
>>
>>
>>
>> l. The standard does not define relation between the u attribute and sharedItems node.
>>
>>
>>
>> In Office, the u attribute applies only when the node parent is sharedItems.
>>
>>
>>
>> Regarding attributes @cp, @f and @u, see discussion above for §18.10.1.2 b.
>>
>>
>>
>> Regarding attributes @bc (Background Color) and @fc (Foreground Color), no default value can be given, because one would expect the defaults in these cases to be controlled by the application.
>>
>>
>>
>> Regarding the attribute @in (Format Index), this clearly only applies if the format string for this entry in the PivotTable is stored in an OLAP serverformat (§18.10.1.86) element.
>>
>>
>>
>> Proposed resolution: Add the default value false to the schema for both @f and @u. Provide an example of the use of @cp on at least one of the six elements on which it is available. No action for the other attributes.
>>
>>
>>
>> §18.10.1.45 item (@x) (complex type: CT_Item)
>>
>>
>>
>> MS-OE376, subsection 2.1.793 (page 353) includes the following notes:
>>
>>
>>
>> k. The standard restricts the x attribute to non-OLAP PivotTables only.
>>
>>
>>
>> Excel allows the x attribute in both OLAP and non-OLAP PivotTables.
>>
>>
>>
>> …
>>
>>
>>
>> m. The standard does not specify when the x attribute is required.
>>
>>
>>
>> Excel requires the x attribute when the value of the t attribute is data.
>>
>>
>>
>> n. The standard states that the x attribute specifies an index to an item in the pivotFields collection.
>>
>>
>>
>> Excel uses the x attribute to specify an index to an item in the sharedItems collection for this pivotField.
>>
>>
>>
>> There no mention of OLAP in the description of the attribute @x for this element in the standard, so note k above can be ignored.
>>
>>
>>
>> Should note m be documented in the standard? The value data is the default value for @t, so it might be useful to document that @x is required if @t is either omitted or has the value data.
>>
>>
>>
>> Note n does not quite agree with the current description of @x in the standard, which states that @x “specifies the item index in pivotFields collection in the PivotCache”. But §18.10.1.67 suggests that a PivotCache in fact contains cacheFields, not pivotFields.
>>
>>
>>
>> Further exploration of the normative text in §18.10 Pivot Tables tends to lead to more confusion. For example, in §18.10.1.46 there is an example which is unintelligible, until one realises that it relates to a longer example in §L.2.9 in the Primer. However, this doesn’t completely resolve the confusion: for example, a value “Adam L Flores” is referred to in §18.10.1.46 and in §L.2.9.3.4.3, but is not included in the example pivotCacheDefinition in §L.2.9.3.2.
>>
>>
>>
>> Proposed action: A thorough review of Pivot Table normative text and Primer material is needed to clarify how to implement Pivot Tables in SpreadsheetML, which might well lead to a need for more wholesale revision of §18.10 and §L.2.9. Tinkering with the description of attribute values here and there will not suffice.
>>
>>
>>
>> §18.10.1.50 m (@bc, @cp, @f, @fc, @in, @u) (complex type: CT_Missing)
>>
>>
>>
>> MS-OE376, subsection 2.1.797 (page 357) includes the following notes:
>>
>>
>>
>> c. The standard does not specify a default value for the f attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> d. The standard does not define the relation between @cp and child elements.
>>
>>
>>
>> In Office, the value of @cp shall be equal to the count of child x elements and be less than or equal to the @mappingCount of the corresponding cacheField element.
>>
>>
>>
>> e. The standard does not define the relation between @cp and the datasource type.
>>
>>
>>
>> In Office, @cp applies only to OLAP-based PivotTables.
>>
>>
>>
>> f. The standard does not define relation between @cp and sharedItems node.
>>
>>
>>
>> In Office, @cp applies only when the node parent is sharedItems.
>>
>>
>>
>> …
>>
>>
>>
>> i. The standard does not specify a default value for the u attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> j. The standard does not define relation between @f and sharedItems node.
>>
>>
>>
>> In Office, @f applies only when the node parent is sharedItems.
>>
>>
>>
>> k. The standard does not define relation between @u and sharedItems node.
>>
>>
>>
>> In Office, @u applies only when the node parent is sharedItems.
>>
>>
>>
>> Regarding attributes @cp, @f and @u, see discussion above for §18.10.1.2 b.
>>
>>
>>
>> Regarding attributes @bc, @fc and @in, see discussion above for §18.10.1.27 e.
>>
>>
>>
>> Proposed resolution: Add the default value false to the schema for both @f and @u. Provide an example of the use of @cp on at least one of the six elements on which it is available. No action for the other attributes.
>>
>>
>>
>> §18.10.1.51 map (@dimension, @measureGroup) (complex type: CT_MeasureDimensionMap)
>>
>>
>>
>> MS-OE376, subsection 2.1.798 (page 358) includes the following notes:
>>
>>
>>
>> a. The standard does not place any restrictions on the value of the measureGroup attribute.
>>
>>
>>
>> In Office, the value of the measureGroup attribute must be less than the number of items in measureGroups collection.
>>
>>
>>
>> b. The standard does not restrict the value of @dimension.
>>
>>
>>
>> In Office, @dimension must be less than number of items the dimensions collection.
>>
>>
>>
>> c. The standard does not specify the base of the dimension attribute index.
>>
>>
>>
>> Office treats the dimension attribute as a zero-based index.
>>
>>
>>
>> d. The standard does not specify the base of the measureGroup attribute index.
>>
>>
>>
>> Office treats the measureGroup attribute as a zero-based index.
>>
>>
>>
>> e. The standard states that the dimension attribute is optional.
>>
>>
>>
>> Excel requires this attribute.
>>
>>
>>
>> f. The standard implies that the measureGroup attribute is optional.
>>
>>
>>
>> Excel requires this attribute.
>>
>>
>>
>> The element has these two attributes and no content. There seems no point in having the element without specifying the two elements that are being mapped/associated: a map and a measureGroup. It would seem to be a mistake in the schema that these attributes are declared to be optional.
>>
>>
>>
>> Proposed resolution: Change the schema to make these two attributes required. No change to the prose.
>>
>>
>>
>> §18.10.1.60 n (@bc, @cp, @f, @fc, @in, @u) (complex type: CT_Number)
>>
>>
>>
>> MS-OE376, subsection 2.1.804 (pages 360–361) includes the following notes:
>>
>>
>>
>> a. The standard does not define relation between @cp and the datasource type.
>>
>>
>>
>> Office allows @cp only with OLAP PivotTables.
>>
>>
>>
>> …
>>
>>
>>
>> e. The standard does not specify a default value for the u attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> f. The standard does not restrict the value of @cp.
>>
>>
>>
>> Office requires that @cp be equal to the count of child x elements and be less than or equal to the mappingCount of the corresponding cacheField element.
>>
>>
>>
>> g. The standard does not specify a default value for the f attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> …
>>
>>
>>
>> n. The standard does not state when @f is valid.
>>
>>
>>
>> In Office, @f is valid only when the parent of the n element is of CT_SharedItems type.
>>
>>
>>
>> o. The standard does not define relation of @u with CT_SharedItems node.
>>
>>
>>
>> In Office, @u is used only when the n element's parent is of CT_SharedItems type.
>>
>>
>>
>> …
>>
>>
>>
>> q. The standard does not place a restriction for on the value of the cp attribute.
>>
>>
>>
>> In Office, the cp attribute applies only when the node parent is sharedItems.
>>
>>
>>
>> Notes f and q appear to answer the same question, but in different ways. Are both correct?
>>
>>
>>
>> Regarding attributes @cp, @f and @u, see discussion above for §18.10.1.2 b.
>>
>>
>>
>> Regarding attributes @bc, @fc and @in, see discussion above for §18.10.1.27 e.
>>
>>
>>
>> Proposed resolution: Add the default value false to the schema for both @f and @u. Provide an example of the use of @cp on at least one of the six elements on which it is available. No action for the other attributes.
>>
>>
>>
>> §18.10.1.74 pivotTableStyleInfo (@showColHeaders, @showColStripes, @showLastColumn, @showRowHeaders, @showRowStripes) (complex type: CT_PivotTableStyle)
>>
>>
>>
>> MS-OE376, subsection 2.1.815 (page 381) includes the following notes:
>>
>>
>>
>> a. The standard defines the @showLastColumn to determine if the last column of formatting is shown.
>>
>>
>>
>> Office ignores this attribute.
>>
>>
>>
>> b. The standard does not specify a default value for the showRowHeaders attribute.
>>
>>
>>
>> Excel uses a default value of 0 for this attribute.
>>
>>
>>
>> c. The standard does not specify a default value for the showRowStripes attribute.
>>
>>
>>
>> Excel uses a default value of 0 for this attribute.
>>
>>
>>
>> d. The standard does not specify a default value for the showColStripes attribute.
>>
>>
>>
>> Excel uses a default value of 0 for this attribute.
>>
>>
>>
>> e. The standard does not specify a default value for the showColHeaders attribute.
>>
>>
>>
>> Excel uses a default value of 0 for this attribute.
>>
>>
>>
>> Given that Excel ignores @showLastColumn, it probably makes sense for the default value to be false.
>>
>>
>>
>> Since all the other attributes have Boolean values, it would make sense for their omission to imply a default value false. One might be able to conceive of a more limited implementation in which a default value true might be appropriate, but this seems unlikely.
>>
>>
>>
>> Proposed resolution: For clarity, add use=“optional” to the schema attribute declarations for @showColHeaders, @showColStripes, @showRowHeaders and @showRowStripes. Add default=“false” to all five attribute declarations.
>>
>>
>>
>> §18.10.1.85 s (@bc, @cp, @f, @fc, @in, @u) (complex type: CT_String)
>>
>>
>>
>> MS-OE376, subsection 2.1.821 (page 385) includes the following notes:
>>
>>
>>
>> a. The standard does not specify a default value for the u attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> b. The standard does not specify a default value for the f attribute.
>>
>>
>>
>> Excel uses a default value of false for this attribute.
>>
>>
>>
>> …
>>
>>
>>
>> f. The standard does not state that @cp is required in certain cases.
>>
>>
>>
>> Office requires that @cp shall be equal to the count of child x elements and be less than or equal to the @mappingCount of the corresponding cacheField element.
>>
>>
>>
>> g. The standard does not state that the presence of @cp is related to the datasource type.
>>
>>
>>
>> In Office, @cp is specific to OLAP-based PivotTables.
>>
>>
>>
>> h. The standard does not define a relationship between @cp and the parent element of s.
>>
>>
>>
>> In Office, @cp applies only when the parent of the s element is of type CT_SharedItems.
>>
>>
>>
>> i. The standard does not define a relationship between @f and the parent element of s.
>>
>>
>>
>> In Office, @f applies only when the parent of the s element is of type CT_SharedItems.
>>
>>
>>
>> j. The standard does not define a relationship between @u and the parent element of s.
>>
>>
>>
>> In Office, @u applies only when the parent of the s element is of type CT_SharedItems.
>>
>>
>>
>> Regarding attributes @cp, @f and @u, see discussion above for §18.10.1.2 b.
>>
>>
>>
>> Regarding attributes @bc, @fc and @in, see discussion above for §18.10.1.27 e.
>>
>>
>>
>> Proposed resolution: Add the default value false to the schema for both @f and @u. Provide an example of the use of @cp on at least one of the six elements on which it is available. No action for the other attributes.
>>
>>
>>
>>
>>
>> ELEMENTS IN DR 16-0007 NOT YET CONSIDERED
>>
>>
>>
>> The following 9 elements in the DR have yet to be considered:
>>
>>
>>
>> §18.10.1.86 serverFormat (@culture, @format) (complex type: CT_ServerFormat)
>>
>>
>>
>> §18.10.1.92 tpl (@fld, @hier) (complex type: CT_Tuple)
>>
>>
>>
>> §18.13.1 connection (@name, @type) (complex type: CT_Connection)
>>
>>
>>
>> §18.13.3 dbPr (@command) (complex type: CT_DbPr)
>>
>>
>>
>> §18.13.5 olapPr (@rowDrillCount) (complex type: CT_OlapPr)
>>
>>
>>
>> §18.13.12 textPr (@characterSet) (complex type: CT_TextPr) – the attribute @codePage was moved to Part 4, but is still mentioned here and in §18.2.24 webPublishing.
>>
>>
>>
>> §18.13.13 webPr (@url) (complex type: CT_WebPr)
>>
>>
>>
>> §18.14.1 cell (@r) (complex type: CT_ExternalCell)
>>
>>
>>
>> §18.14.15 definedName (@refersTo) (complex type: CT_ExternalDefinedName)
>>
>>
>>
>
>
>
> --
>
> Praying for the victims of the Japan Tohoku earthquake
>
> Makoto
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.vse.cz/pipermail/sc34wg4/attachments/20170819/c6736974/attachment-0001.html>
More information about the sc34wg4
mailing list