The diagram we showed in the last post was somewhat misleading. The concluding stage of creating an Asciidoctor document looks more like this:
If you’re doing anything moderately complicated, your document may not end up looking like what you expect. Even if your document is not that complicated, you may end up with an error in markup that causes Metanorma to complain, or to crash. And even if those things do not happen, the syntax of the document you have authored may violate the requirements of the standard body, as captured in the formal grammar used inside Metanorma.
To deal with this, you need to run Metanorma over your Asciidoctor input frequently, as you build up its content and markup, so you can narrow down where anything went wrong to what you have updated most recently. If that doesn’t work in narrowing it down, you need to cut the document into smaller and smaller pieces, until you have identified the bit of the document causing the problem. By gradually building up the document, and testing it as you go, you can prevent unpleasant surprises later on.
The fix depends on what has gone wrong, of course; some of the first things to be careful of are:
Blank lines need to be used to set blocks of text off from each other. Asciidoctor marks up paragraph-level objects (blocks), such as tables, lists, and notes: these must be separate by a blank line from any preceding paragraphs, otherwise they will be treated as part of that paragraph, and the markup that introduces them will be ignored.
Characters that are being misinterpreted as markup — such as square brackets at the start of a line.
The right level is applied to each heading (as indicated by the number of equals signs before it).
An initial markup tag is matched by its ending tag; for example, if
_(underscore) starts an italicized passage, make sure the passage also contains the ending
_. If the initial tag involves multiple characters as a delimiter (we’ll come to those), its corresponding ending tag must have the same number of characters.
Any crossreferences in the text (we’ll come to those too) are correct.
Markup should be kept within a line of text, and not be interrupted by a carriage return. So if an italicized passage spans more than one line, close the italics at the end of each line, and restart it at the beginning of the next. If that is not possible, particularly for any markup contained within a pair of square brackets, you will just need to keep the marked up content as one long line.
The grammar warnings that Metanorma issues about violating the expected structure of the document do also need to be looked at, at least once. However, the structure of the document needs to be complete for it to be valid; so you can defer resolving any grammar issues until you have finished putting the document together.
Checking the grammar involves you hopping between the Asciidoctor input, and the XML intermediate form (which is what is parsed for grammatical issues). Checking that the document works as expected involves you hopping between the Asciidoctor input, and one of the outputs, either HTML, PDF, or DOC. (The HTML is easier to work with for error tracing than the DOC; but eventually, you will have to check both.)
Levels of markup
Asciidoctor has three levels of markup:
sections, such as clauses and subclauses, marked up by headings;
blocks, which are paragraphs and spans with the same scope as paragraphs, like lists and tables;
and inline markup, such as italics and hyperlinks, which is contained within a block.
Asciidoctor keeps these three levels distinct. You can’t have italics markup span across two blocks, for example. (As we saw, bad things can happen to some inline markup if it even spans across two lines.) A block cannot contain another block; that’s why there needs to be a blank line between a paragraph and a list (although lists can have multiple levels of nesting.) And every block is meant to be contained in a section — except for blocks before the first section header, which defines a section of its own (the preamble, in Asciidoctor terms; Metanorma uses it as the foreword).
We have seen that the document can have attributes, defined in the document header. Sections and blocks can also have attributes; they are indicated by a line contained in square brackets before the section header or block, which contains the attribute values. Blocks can also have captions (a line preceded by period/full stop). Those two features are why a normal line of text can’t start with an open square bracket or dot. And both blocks and sections can have an anchor, included in double square brackets: that is the identifier used to cross-reference the block or section from elsewhere in the text.
Example: Metanorma table in Asciidoctor
For example, the following is an Asciidoctor table:
[[tableD-1]] [cols="<,^,^,^,^",headerrows=2] .Repeatability and reproducibility of husked rice yield |=== .2+| Description 4+| Rice sample | Arborio | Drago footnote:[Parboiled rice.] | Balilla | Thaibonnet | Number of laboratories retained after eliminating outliers | 13 | 11 | 13 | 13 | Mean value, g/100 g | 81,2 | 82,0 | 81,8 | 77,7 |===
The first line is the anchor for the table: if you want to refer to the table
from elsewhere in the document, you will use the Asciidoctor inline markup for
<<tableD-1>>. The third line is a caption for the table.
Rendering Metanorma will automatically replace the anchor and caption with
an auto-numbered caption for the table
(e.g. Table 3. Repeatability and reproducibility of husked rice yield);
if you add a table before this table and regenerate the Metanorma outputs,
the caption number will automatically update. More critically, the crossreference
to that table will be automatically replaced with an auto-numbered reference,
such as Table 3, which again will be updated automatically if you insert more
The second line is a set of attributes of the table; they include an attribute on how each column will be text aligned (first left-aligned, the remainder center-aligned), followed by a Metanorma-specific attribute, indicating that there are two header rows in the table. (Asciidoctor proper only supports one.)
The table itself, as is typical of blocks, has a starting delimiter,
indicating where this particular block starts, and a matching end delimiter,
|===, both on a separate line. Between the two delimiters, there is table-specific
markup: as you can guess,
| indicates a new cell, while
4+ indicates that
the following cell spans four columns, and
.2+ that the following cell
spans two rows.
|This table example originates from the IsoDoc Rice document, which demonstrates document elements used in ISO International Standards.|
Markup on sections looks similar.
Example: Metanorma bibliography in Asciidoctor
The following is how a bibliography may be marked up:
[[infobib]] [bibliography,obligation=infomrative] == Informative Bibliography * [[[ISO3696,ISO 3696]]], _Water for analytical laboratory use -- Specification and test methods_ == Next heading...
Again, the first line is a crossreference anchor, and the third line is a title (although this time it is a heading, and a caption, because this is a section and not a block.)
The second line contains attributes, including the style attribute
indicating what kind of a section this is. The style attribute is critical for the sections
that need it: in this case, the triple bracket notation used to define references
is only processed if it is contained in a section with a
bibliography style attribute.
There is no distinct delimiter for sections — or rather, the heading is what delimits
the start of the section, and the next heading (or end of document) delimits its end.
|In the Metanorma toolchain cited documents from certain standards bodies, such as ISO, IETF and IEC can be auto-fetched, meaning that the title doesn’t need to be supplied, just like this:|
* [[[ISO3696,ISO 3696]]], _AUTOFILLED BY METANORMA_
This information will help you go back over the document you are marking up, and give it the structure it needs to make sense as a standards document.