Document is not preserved through load and save

[JackNWhite]

I have a barebones document to which I would like to add new sections. I am simply loading and then saving the document like so:

let template = Document::open("front_page.docx")?;
let result = template.save("translated.docx");

I have the following problems:

• Two custom XML fields which are deliberately duplicated - on the front page and in the general footer - are not present on the front page
• The style of tables is not preserved
• Two tables are removed entirely and replaced with their contents as plain text
• The tables of figures and tables lose their fields altogether and are replaced with the text of the field code, e.g. TOC \h \z \c "Figure"No table of figures entries found.

Is there a way to clear these issues up?

[mantissaman]

Thanks for the clear report. You found two separate bugs, and both are real. Here is where each one stands.

The tables of figures losing their fields. This is fixed on main. A field like a TOC is written as five runs, and three of them hold only a w:fldChar marker for begin, separate and end. Those markers are empty XML elements, and the run parser dropped every empty element it did not recognise. The w:instrText in the middle survived, so what you got back was the field instruction with no field around it, which Word then shows as ordinary text. That is exactly your TOC \h \z \c "Figure"No table of figures entries found. The fix landed in #5 earlier today.

The tables being removed. This one was still live and I have a fix in #9. CT_Tbl, CT_Row and CT_Tc parsed the children they knew about and threw the rest away. Everywhere else in the tree we keep unknown XML verbatim, so tables were the one place a plain load and save deleted content. Round tripping on main today:

input	what came back
a row inside a content control	the whole row gone
a cell inside a content control	a row with no cells
a paragraph inside a content control	a cell with no paragraph
a bookmark at table or row level	gone

Content controls around rows are common in template documents, which fits the front page you describe. Bookmarks going missing would also break any cross reference pointing into those tables. All of these round trip byte for byte with #9 applied.

The two custom XML fields on the front page. I suspect this is the same content control bug, because a field on a front page is usually a w:sdt and it is often inside a layout table. If your front page has the field inside a table, #9 should bring it back. Worth retesting once that lands.

The table styles. I could not reproduce this one. A w:tblStyle inside w:tblPr round trips fine on its own here, so something more specific is going on. Could you attach the file, or a cut down version of it? If it cannot be shared publicly then a snippet of the w:tblPr block from word/document.xml for one of the affected tables would give me enough to work with.

Sorry this bit you on such a basic path. A load and save with no edits should give the document back unchanged, and it did not.

[mantissaman]

Update on all four of these.

Tables of figures losing their fields. Fixed in #5. The w:fldChar markers that open and close a field are empty XML elements and the run parser dropped every empty element it did not recognise, so the field instruction survived with no field around it and Word showed it as ordinary text.

Tables removed and replaced with plain text. Fixed in #9. CT_Tbl, CT_Row and CT_Tc threw away any child they did not recognise, so a row, cell or paragraph wrapped in a content control was deleted outright. Bookmarks at table and row level were going the same way.

The style of tables. Fixed now, and I owe you a correction. I said earlier I could not reproduce it. What I had actually checked was that w:tblStyle round trips, which it does, so it looked clean. The style reference was never the problem.

What was being lost is the markup that says which parts of that style to apply:

IN :  <w:tblLook w:val="04A0" w:firstRow="1" w:lastRow="0" w:firstColumn="1" w:noVBand="1"/>
OUT:  <w:tblLook w:val="04A0"/>

IN :  <w:trPr><w:cnfStyle w:val="100000000000"/></w:trPr>
OUT:  gone

w:tblLook kept only its w:val and w:cnfStyle was not modelled at all. Those are what turn on the header row emphasis, the row banding and the first column formatting. Without them Word keeps the style name and draws the table with base formatting only, which is exactly what you described. Both round trip properly now.

The two custom XML fields on the front page. This is the one I still cannot confirm. My guess is that they are content controls inside a layout table, in which case #9 already fixes them, but that is a guess and I would rather not close it on one.

Could you retest against 0.4.0 once it is out? If the fields are still missing, the most useful thing would be the word/document.xml fragment around one of them, or a cut down file that still loses it.

Sorry this took as long as it did. A load and save with no edits should give the document back unchanged, and on four separate counts it did not.