Organization for the Advancement of Structured Information Standards (OASIS) Technical Memorandum TR 9901:1999 NormanWalsh Chair, Tables Technical CommitteeArbortext, Inc. 1999 Sep 29 1999Organization for the Advancement of Structured Information Standards (OASIS) Permission to reproduce parts or all of this information in any form is granted to OASIS members provided that this information by itself is not sold for profit and that OASIS is credited as the author of this information. This OASIS Technical Memorandum is an XML expression of the Exchange subset of the full CALS table model DTD described in OASIS Technical Memorandum 9502:1995, CALS Table Model Document Type Definition. It is an XML version of OASIS Technical Resolution 9503:1995, Exchange Table Model Document Type Definition. The Exchange subset has been chosen as being a useful subset of the complete CALS table model such that, if an application's tables are tagged according to this subset, there is a high probability that the table will be interoperable among the great majority of OASIS vendor products. See also OASIS Technical Research Paper TRP 9501:1995, Table Interoperability: Issues for the CALS Table Model. Technical Memorandum 9901:1999XML Exchange Table Model 1.0 1999 Sep 30 nwalsh First publication. 0.95 1999 Aug 29 nwalsh Revised draft. TR has become TM, %tbl.table.att documented, default declaration for %bodyatt added to DTD. 0.9 1999 Apr 15 nwalsh Committee draft 0.5 1999 Apr 10 nwalsh Editorial draft

This OASIS Technical Memorandum is an XML expression of the Exchange subset of the full CALS table model DTD described in OASIS Technical Memorandum 9502:1995, CALS Table Model Document Type Definition. It is an XML version of OASIS Technical Resolution 9503:1995, Exchange Table Model Document Type Definition. The Exchange subset has been chosen as being a useful subset of the complete CALS table model such that, if an application's tables are tagged according to this subset, there is a high probability that the table will be interoperable among the great majority of OASIS vendor products. See also OASIS Technical Research Paper TRP 9501:1995, Table Interoperability: Issues for the CALS Table Model. Note that the set of element and attribute declarations in the Exchange Table Model DTD Module partially defines the Exchange table model. However, the model is not well-defined without the accompanying natural language description of the semantics (meanings) of these various elements, attributes, and attribute values. The semantic writeup, in the sections following the markup declaration module, must be used in conjunction with the element and attribute declarations. The following markup declaration module defines parameter entities for various parts of the model. This organization provides for more flexibility in using this module in different circumstances. However, it must be realized that overriding declarations of these parameter entities that enlarge the model (i.e., that define a superset of the model or a different model) may reduce practical interoperability. It is therefore recommended that the parameter entities be used primarily to define a subset of the default model. Any parameter entity redefinition other than to define a more constrained model than the default model must be done carefully with the knowledge of ones target toolset's capabilities and with the realization that interoperability may be reduced. In particular, for maximum interoperability, the geometric aspects, including some table attributes, and the elements and attributes of the table structure: tgroup through row and some entry attributes should only be constrained. The table content model and the entry content model are the primary candidates for change other than constraint.

Specific changes of substance between this Exchange Model and the SGML Exchange Model described in OASIS Technical Resolution TR 9503:1995 include: Removed the tbl.table.excep, tbl.hdft.excep, tbl.row.excep, and tbl.entry.excep parameter entities. There are no exceptions in XML. The following normative statement is made in lieu of exceptions: the exchange table model explicitly forbids a table from occurring within another table. If the content model of an entry includes a table element, then this cannot be enforced by the DTD, but it is a deviation from the exchange table model to include a table within a table. Removed the tbl.hdft.name, tbl.hdft.mdl, tbl.hdft.excep, and tbl.hdft.att parameter entities. The motivation for these elements was to change the table header/footer elements. Since XML does not allow element declarations to contain name groups, and the exchange table model does not allow a table to contain footers, the continued presence of these attributes is unnecessary. Added the tbl.thead.att parameter entity. This entity parameterizes the attributes on thead. It replaces the tbl.hdft.att parameter entity. Tag ommission indicators have been removed. Comments have been removed from declarations. NUMBER attributes have been changed to NMTOKEN. NUTOKEN attributes have been to changed to NMTOKEN. Removed the grouping characters around the content model parameter entry for the entry element. This is necessary so that an entry can contain #PCDATA and be defined as an optional, repeatable OR group beginning with #PCDATA, as required by XML.

Specific changes of substance between this Exchange Model and the CALS table model contained in the Example DTD published in MIL-HDBK-28001 (30 June 95) include: change most explicit attribute defaults from a value to #IMPLIED to allow value inheritance, possibly from a style sheet; eliminate entrytbl; eliminate tfoot; eliminate spanspec; eliminate the char and charoff attributes from tgroup; eliminate the rotate attribute from entry; eliminate references to security attributes. Finally, this declaration set is assigned a Formal Public Identifier.

Note that the above set of element and attribute declarations in the previous section only partially defines the Exchange table model. The model is not well-defined without the accompanying natural language description of the semantics (meanings) of these various elements, attributes, and attribute values. This section containing that semantic writeup must be used in conjunction with the element and attribute declarations.

The table element markup identifies a table. Elements inside a table may at times inherit default values from the attributes on the containing table. Furthermore, a table element may have some stylesheet associated with it that may provide default values for some or all of its attributes. If a table element has no explicit specification for an attribute but does have an associated style sheet that gives a specification for this attribute, then the stylesheet value shall be used as the value that is inherited from this element.

Describes position of outer rulings. Declared value The enumerated values are: ValueRules sides left and right sides top above first row bottom after last row topbot top and bottom all all of the above none none of the above The outer rulings appear in place of and in the space that would otherwise be taken by horizontal and vertical rulings on the outsides of those entries that appear at the edges of the table. Default IMPLIED (implies value from stylesheet if available, if not, implies “all”).

Specifies the presence or absence of column separator rules. Provides the default value for all tgroups in this table. If colsep is non-zero, display internal column rules to the right of each entry; if zero, do not display the rules. This value is ignored for the rightmost column, where the frame setting applies. Declared value %yesorno; (NMTOKEN) Default IMPLIED (implies value from stylesheet if available, if not, implies “1”).

Specifies the presence or absence of row separator rules. Provides the default value for all tgroups in this table. If rowsep is non-zero, display the internal row rules below each entry; if zero, do not display the rules. Ignored for the last row of the table (i.e., the last row of the last tgroup in this table), where the frame value applies. Declared value %yesorno; (NMTOKEN) Default IMPLIED (implies value from stylesheet if available, if not, implies “1”).

Describes the desired width of the table. If zero, the maximum available width for the table is the (galley) width (possibly respecting current indents in force as specified by the stylesheet) of the current column of page. If non-zero, the table spans the width of the entire page (possibly causing any previous multicolumn text on the page to be balanced and any extra processing associated with column balancing and page spanning to be performed). Unlike most table attributes, this one is parameterized by %tbl.table.att;and may be removed by a customization layer. Declared value %yesorno; (NMTOKEN) Default IMPLIED (implies value from the stylesheet if available). In the absence of an explicit value (or one implied from the stylesheet), the system should attempt to format the table into a galley column if reasonable, based on explicit colspec colwidth values, galley column width, and possible table indentation implied by context. If any relative width specifications exist, they should continue to apply, but the unit proportion should be first based on the width uncommitted from the galley column. In the preceding paragraph, “if reasonable” is intended to preclude a system generating columns so narrow that the entry content is obscured by awkward line folding, or clipping. For character alignment, the content on either side of the alignment character should also be considered. No explicit algorithm for determining when a table will fit is provided. Because of the uncertainty in meaning, interchange of a table should include an explicit value for pgwide.

Any of the attributes in the associated attribute set may be used with this element. Default As appropriate for each attribute in the set. This parameter entity is included for backward compatibility with the SGML exchange table model. The XML table model also defines %tbl.table.att; for this purpose, parallel to the other parameter entities added for defining attributes on table elements.

Any of the attributes in the associated attribute set may be used with this element. Default As appropriate for each attribute in the set.

Each tgroup effectively identifies a new portion of a table. The colspecs and the colnames in each tgroup are independent of any other tgroup. The colspecs in a tgroup apply to its tbody. The colspecs in a tgroup shall be in left-to-right column order. Explicit colnums in the different colspecs of a tgroup shall be unique, in ascending order left-to-right, in the range 1 to the number of colspecs, which is not to exceed the value of the cols attribute of tgroup. The colnames in the different colspecs of a tgroup shall be unique. (A colname may have the same value as the colnum of the same colspec if restricted to digits.) The colspecs of tbody also apply to thead. All tgroups of a table shall have the same width, so the table frame can surround them uniformly. Each thead and tbody has that total width as well. For each tgroup, let the minimum tgroup width be the sum of the fixed portions of the set of colspec colwidths in that tgroup which should not exceed the indicated total width, determined from the table pgwide attribute. If pgwide is "0", the target total width is the galley column width, possibly reduced by the current indents in force; if pgwide is non-zero, the target total width is the full page width. The minimum table width shall be the maximum of all the minimum tgroup widths. The total width for any tgroup may be specified with some columns having fixed colwidths and others having proportional widths. When no proportional width are specified for a given tgroup, the width of that tgroup shall be the sum of the fixed widths of the columns. When any proportional widths are specified, the unit proportion is determined by starting with the target table width determined from the table pgwide value non-zero for full width, “0” for galley column width (possibly affected by current indents). Reduce that by the sum of border widths and column ruling widths and the fixed colwidths to get the available proportional width. Sum the proportional factors N over all columns with proportional parts “N*” and including “1*” for each implicit colwidth when the value of the tgroup cols exceeds the number of colspecs in that tgroup. Then the unit proportion is the available proportional width divided by the sum of the proportionality factors. Note that the fixed widths for frame and colsep rulings are deemed to be small and system dependent, not essentially included as fixed parts of each colwidth in colspec. If the table fixed widths take up too much of the indicated table width, then it is up to the output system to resolve the conflict. If there are multiple tgroups in a single table with fixed widths (i.e., no proportional parts) such that these fixed widths differ from one another, then it is up to the output system to determine the interpretation. Elements inside a tgroup may at times inherit default values from the attributes on the containing tgroup. Furthermore, a tgroup element may have some stylesheet associated with it that may provide default values for some or all of its attributes. If a tgroup element has no explicit specification for an attribute but does have an associated style sheet that gives a specification for this attribute, then the stylesheet value shall be used as the value that is inherited from this element.

Number of columns in the tgroup. The value of this attribute must be an integer greater than zero. Declared value NMTOKEN

Provides default for all colspecs in this tgroup. If other than zero, display the internal column rulings to the right of entry; if zero, do not display them. Ignored for the last (rightmost) column, where the frame setting applies. Possible default source for colspec or entry. Declared value %yesorno; (NMTOKEN) Default IMPLIED (implies value from stylesheet if any, else from table)

Provides default for colspec s in this tgroup. If other than zero, display the internal horizontal row ruling below each entry. If zero, do not display them. Ignored for the last (bottom) row of the table where the frame value applies. In the last row in any tgroup other than the last (or only) in the table, the regular row or entry rowsep specifications apply. Possible default source for row or entry. Declared value %yesorno; (NMTOKEN) Default IMPLIED (implies value from stylesheet if any, else from table)

Text horizontal position within the column. Applies to text that is #PCDATA or other in-line elements, not further contained in another element with its own formatting style, such as paragraph, list, or annotation. Default source for colspec align. Possible default source for entry. Declared value The enumerated values are: ValueAlignment left Flush left center Centered right Flush right justify Flush left and right char Align text to the leftmost occurrence of the value of the non-null attribute char value. Default IMPLIED (means use value from stylesheet if any, else “left”.)

Specifies a column, a vertical portion of a table. The default values come from the tgroup or thead starting the current (enclosing) group. Each colspec is for a single column in left-to-right order, so it properly has a column number, colnum, implicitly in order starting from 1, and an optional column name (colname) by which it is known when referenced by any entry . Any references from entrys within the thead to colname, namest, or nameend refer to values defined by the set of colspecs in the containing tgroup. Colspecs from the containing tgroup apply to thead and tbody. The number of columns should be determined by the cols attribute on the tgroup element, not by the number of colspecs actually defined. If the number in cols is larger than the number of colspecs, then additional colspecs of colwidth “1*” should be inferred. Colspecs can be numbered or unnumbered, and if numbered should be increasing in sequential order starting with 1 at the left. Unnumbered colspecs are interpreted as being numbered incrementally (one more than the previous column number), with the first colspec starting at 1. It is left up to the implementation how to handle any mismatch in colspec numbering, or a number of colspecs greater than cols. It is recommended that an authoring or editing implementation or any implementation that verifies the compliance of the table markup to this Memorandum offer the option of producing a warning message when it encounters such markup. It is an error for an authoring or editing implementation to produce a table with such conflicting markup. The char and charoff attributes are allowed on colspec as well as on entry. While an implementation should accept these attributes on either of these elements upon input, it may produce logically equivalent output that nevertheless differs with respect to these attributes and still be considered in compliance with the semantics of this Memorandum. If a table contains one or more entrys with values of either char or charoff that differ from the value for the same attribute on a colspec that logically contains the entry (s), it is left up to the implementation how this should be handled (i.e., the interpretation of such markup is beyond the scope of this model). It is recommended that an authoring or editing implementation or any implementation that verifies the compliance of the table markup to this Memorandum offer the option of producing a warning message when it encounters such markup. It is an error for an authoring or editing implementation to produce a table with such conflicting markup.

Number of column, counting from 1 at left of the table. The value of colnum is not useful to identify a column in an entry, so serves no functional purpose other than a consistency check on the order of the colspecs. Declared value NMTOKEN Default IMPLIED (colspec is the next one in order)

Name of column, used to specify the position or horizontal span of columns in a row by reference in entry using colname, namest, and/or nameend. The colname value could be the same as colnum, as its declared value is NMTOKEN, though there is no such requirement. The name space for colnames is different for each tgroup. Declared value NMTOKEN Default IMPLIED (colspec cannot be referenced without colname)

Either proportional measure of the form number*, e.g., “5*” for 5 times the proportion, or “*” (which is equivalent to “1*”); fixed measure, e.g., 2pt for 2 point, 3pi for 3 pica. (Mixed measure, e.g., 2*+3pt, while allowed in the full CALS table model, is not supported in this Exchange model.) Coefficients are positive integers or fixed point numbers; for fixed point numbers, a leading (possibly 0) integer part is required, and implementations should support at least 2 decimal places. A value of "" [the null string] is treated as a proportional measure of “1*”. The fixed unit values are case insensitive. The standard list of allowed unit values is “pt” (points), “cm” (centimeters), “mm” (millimeters), “pi” (picas), and “in” (inches). The default fixed unit should be interpreted as “pt” if neither a proportion nor a fixed unit is specified. Declared value CDATA Default IMPLIED (means assume a proportional measure of “1*”)

Default for column ruling to the right of entrys starting in this column (within the tgroup). Declared value %yesorno; (NMTOKEN) Default IMPLIED, from tgroup.

Default for row ruling below an entry starting in this column when there is neither a rowsep value on the entry nor on the row in which the entry occurs. Note that with a non-zero value for morerows on the entry, that ruling will be several rows lower. Declared value %yesorno; (NMTOKEN) Default IMPLIED, from tgroup.

Text horizontal position within the column or spanning columns. Possible default source for entry. Declared value The enumerated values are: ValueAlignment left Flush left center Centered right Flush right justify Flush left and right char Align text to the leftmost occurrence of the value of the non-null attribute char value. Default IMPLIED (means use value from tgroup)

Default source for entrys starting in this column. If character alignment is specified, the value is the single alignment character source for any implied char values for entry immediately in this column. A value of "" (the null string) means there is no aligning character. Declared value CDATA Default IMPLIED, means "" (i.e., there is no aligning character).

Default source for entrys starting in this column. For character alignment on an entry in the column, horizontal character offset is the percent of the current column width to the left of the (left edge of the) alignment character. This value should be number, greater than 0 and less than or equal to 100. Declared value NMTOKEN Default IMPLIED, means "50" (i.e., 50%)

Identifies the heading rows of a tgroup, displayed as the first rows, and again at the top of any continuation after a physical break between rows in tbody. The attributes apply from the set of colspecs of the containing tgroup .

Default text vertical positioning within the entrys. Provides default value for rows and entrys in thead. Declared value The enumerated values are: ValueAlignment top top middle approximately vertically centered bottom bottom Default IMPLIED (implies “bottom”).

Identifies the body of a tgroup.

Default Text vertical positioning within the entrys. Provides default value for row and entrys in tbody. Declared value The enumerated values are: ValueAlignment top top middle approximately vertically centered bottom bottom Default IMPLIED (implies “top”)

Identifies the row information in a thead or tbody element. Default values come from the enclosing thead, tbody, or tgroup attribute values for like-named attributes. The number of columns consumed by the entrys in a row including their spans, and by columns encroached by an entry with morerows from a prior row of a tgroup, shall not exceed the cols attribute value in effect for this tgroup. It is left up to the implementation how to handle the situation when more than one entry would fill any column of a row, including by spanning or straddling. It is recommended that an authoring or editing implementation or any implementation that verifies the compliance of the table markup to this Memorandum offer the option of producing a warning message when it encounters such markup. It is an error for an authoring or editing implementation to produce a table with such conflicting markup. If the number of columns consumed is less than the number of columns specified in the tgroup's cols attribute, then the missing entrys (without spanning or straddling) are implicitly present, and their colsep and rowsep rulings are inherited.

Default for all entrys starting in this row that do not specify rowsep. If other than zero, display the internal horizontal row ruling below an entry in the row. If zero, do not display it. Rowsep is ignored for the last row of the table where the frame specification determines the ruling. Declared value %yesorno; (NMTOKEN) Default IMPLIED, from the closest element in the inheritance path: tgroup then table.

Text vertical positioning default for entrys in a row. Declared value The enumerated values are: ValueAlignment top top middle approximately vertically centered bottom bottom Default IMPLIED, from the closest element in the inheritance path: either thead or tbody, whichever is the parent of this row.

Identifies an entry in a row. Default values for colsep, rowsep, valign, align, char, and charoff may come from like-named attributes from the nearest element identified as influencing this entry. The values may come from table, tgroup, colspec (defining namest or colname), tbody, or row. Note that colspec is not strictly in the element lineage. The char and charoff attributes are allowed on colspec as well as on entry. While an implementation should accept these attributes on either of these elements upon input, it may produce logically equivalent output that nevertheless differs with respect to these attributes and still be considered in compliance with the semantics of this Memorandum. A table with one or more entrys with values of either char or charoff that differ from the value for the same attribute on any other entry in that logical column or from a colspec that logically contains the entry(s) does not conform to the model defined in this Memorandum. It is left up to the implementation receiving such a table to determine how it should be handled. It is recommended that an authoring or editing implementation or any implementation that verifies the compliance of the table markup to this Memorandum offer the option of producing a warning message when it encounters such markup. An entry gets its defaults from its starting column. A row has no explicit entry in any column into which a vertical straddle occurs because of a morerows attribute on an entry from a prior row. A spanning entry has no explicit repeated entrys in the columns to the right of the column identified by namest. An entry without specific starting column (a namest attribute) cannot span, and falls in the next non-straddled and non-spanned column. The precedence for determining the column(s) for any entry is: span before individual column: namest through nameend individual column: namest colname if no namest implicit if neither namest or colname The implicit Memorandum places the entry in the next available column. There are various markup combinations which this Memorandum defines as “erroneous” and referred to as “errors” in some of the following attribute descriptions. It is left up to the implementation how to handle each of these situations: any column referenced by name (either colname, namest or nameend) is not a colname of a colspec in that tgroup; any of those names are multiply defined in different colspecs of that tgroup; any portions of different entrys overlap each other; morerows vertical straddling extends beyond the rows appropriate to that thead or tbody; too many entrys occur in any row, exceeding the cols on the tgroup; any colnames used in namest and nameend to define a span in an entry refers to the namest column that is not to the left of the nameend column; an entry specifies a start column via entry's namest or colname that is to the left of the column where the entry would be placed by default. It is recommended that an authoring or editing implementation or any implementation that verifies the compliance of the table markup to this Memorandum offer the option of producing a warning message when it encounters such markup. It is an error for an authoring or editing implementation to produce a table with such markup. If the content of an entry is too deep for the available depth of a page, what information is presented and how it is presented is left to be determined by the implementation.

Column name of entry. Ignore if namest is present. It is an error if colname is not defined in a colspec of the current tgroup. Declared value NMTOKEN Default IMPLIED (column(s) determined by namest or implicitly the next.)

Name of leftmost column of span. The value must be some colname in a colspec of the current tgroup. A namest on an entry with no nameend specification indicates the single column so named, though colname is usually used for such a specification. If neither namest nor colname occur, the entry goes in the next column in sequence to the right, skipping over any encroachment from a prior row via entry morerows=N. It is an error if the namest value is not defined in a colspec for the current tgroup. Declared value NMTOKEN Default IMPLIED (implies this attribute has no effect on spanning)

Name of rightmost column of span. The value must be some colname in a colspec of the current tgroup. The column must be to the right of the column identified by namest. A nameend attribute without a namest is ignored. Declared value NMTOKEN Default IMPLIED (implies this attribute has no effect on spanning)

There shall be at least that many more rows in the appropriate thead or tbody. Any entries with morerows that would attempt to extend further downward is an error. The rowsep is determined when the entry is processed, in the top row if morerows is positive. Declared value NUMBER Default IMPLIED (interpret as “0”)

If other than zero, display the internal vertical column ruling at the right of the entry; if zero, do not display it. Ignored for the last column of a row, where the frame setting applies. The colsep value is determined by the leftmost column of a spanning entry, even though its effect is on the last column of such spanning entry. Declared value %yesorno; (NMTOKEN) Default IMPLIED, from the closest element in the inheritance path: colspec then tgroup then table. Note that if a tgroup or table element has no explicit specification for this attribute but does have an associated style sheet that gives a specification for this attribute, then this value shall be used as the value that is inherited from this element.

If other than zero, display the internal horizontal row ruling below the entry; if zero, do not display it. Ignored for the last row of tgroup where the table frame applies. Rowsep for the entry is determined in the topmost row in which the entry occurs, even though its effect may be morerows below. Declared value %yesorno; (NMTOKEN) Default IMPLIED, from the closest element in the inheritance path: row then colspec then tgroup then table. Note that if a tgroup or table element has no explicit specification for this attribute but does have an associated style sheet that gives a specification for this attribute, then this value shall be used as the value that is inherited from this element.

Text horizontal position within the column or spanning columns. Applies to text that is #PCDATA or other inline elements not further contained in another element with its own formatting style, such as paragraph, list, or annotation though such elements could inherit their alignment from this value. Declared value The enumerated values are: ValueAlignment left Flush left center Centered right Flush right justify Flush left and right char Align text to the leftmost occurrence of the value of the non-null attribute char value. Default IMPLIED (means use value from nearest applicable colspec of the nearest ancestral tgroup. If none, then “left”.)

If character alignment is specified, the value of the char attribute is the single alignment character on which the first to occur of this character in the entry is aligned. A value of "" (the null string) means there is no alignment character. If there is no alignment character or the alignment character does not occur in the entry, the entry right aligns to the left of the charoff position. Declared value CDATA Default IMPLIED, from colspec else there is no aligning character.

When character alignment is specified for this entry, horizontal character offset is the percent of the current column width to the left of the (left edge of the) alignment character. This value should be number, greater than 0 and less than or equal to 100. Declared value NMTOKEN Default IMPLIED, from colspec else "50" (i.e., 50%).

Text vertical positioning within the entry. Declared value The enumerated values are: ValueAlignment top top middle approximately vertically centered bottom bottom Default IMPLIED from the closest ancestor with explicit specification of valign: row then thead or tbody.

span A spanning table cell is one that occupies more than one column. In this DTD, the only way a table cell (entry) can span is by specifying both namest and nameend attributes. Spanning and straddling are independent; a spanning cell can also straddle. straddle A straddling table cell is one that occupies more than one row. In this DTD, the only way a table cell (entry) can straddle multiple rows is by specifying a morerows attribute. Spanning and straddling are independent; a spanning cell can also straddle. The following contributed to the development of this Memorandum: Harvey Bingham Carla Corkern (ISOGEN) Tony Graham (Mulberry Technologies) Paul Grosso (Arbortext) Eduardo Gutentag (Sun Microsystems) Ken Hollman (Crane Softwrights) Eve Maler (Arbortext) Norbert Mikula (DataChannel) Nam Jin Son (ISOGEN) Marcy Thompson (ISOGEN) B. Tommie Usdin (Mulberry Technologies) Norman Walsh, Chair (Arbortext)