feat(ods): a column's default cell style reaches its cells

docs/PARITY.md

@@ -567,6 +567,14 @@ The reader opens `content.xml` and `meta.xml`. It does not open
 default cell styles and all page setup — so a LibreOffice-authored file
 reads back with its direct formatting only.
 
+A column's `table:default-cell-style-name` _is_ read, and fills in for
+cells that name no style of their own. That matters because LibreOffice
+puts a column's number format there rather than on the cells: without it
+a LibreOffice document came back with its values and none of its formats.
+A column pointing at a named style — `"Default"`, which it writes on the
+columns it did not format — still resolves to nothing, for the reason
+above.
+
 See [What ODS carries](../README.md#what-ods-carries) for the consequences
 worth knowing before relying on it.
 

docs/SPEC-COVERAGE.md

@@ -51,7 +51,7 @@ Corpus under `test/fixtures`: 35 OOXML workbooks (455 element names, 245 attribu
 
 ## OpenDocument (OASIS ODF 1.3)
 
-153 elements and 224 attributes in the spreadsheet-relevant namespaces. 116 are named somewhere in `src/`; 261 are not.
+153 elements and 224 attributes in the spreadsheet-relevant namespaces. 117 are named somewhere in `src/`; 260 are not.
 
 ### In an ODF document here, not yet looked at
 
@@ -61,19 +61,18 @@ office:settings
 
 ### In an ODF document here, looked at and left
 
-| name                            | why                                                                                                                                                                                                  |
-| ------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `number:boolean-style`          | a data style for booleans. Excel has no format code for one — it shows TRUE/FALSE — so there is nothing to map it to                                                                                 |
-| `number:fill-character`         | the character Excel's `*` repeats to pad a cell. `*` is not in hucre's format model at either end                                                                                                    |
-| `table:automatic-find-labels`   | formula label lookup — no formula engine                                                                                                                                                             |
-| `table:calculation-settings`    | calculation preferences — hucre has no formula engine                                                                                                                                                |
-| `table:case-sensitive`          | formula comparison rule — no formula engine                                                                                                                                                          |
-| `table:default-cell-style-name` | **open** — a column's default cell style. hucre reads direct formatting only, and `PARITY.md` records that the ODS reader does not open `styles.xml`; this is the `content.xml` half of the same gap |
-| `table:iteration`               | iterative calculation settings — no formula engine                                                                                                                                                   |
-| `table:maximum-difference`      | iteration convergence bound — no formula engine                                                                                                                                                      |
-| `table:null-year`               | the century a two-digit year resolves into, for the formula parser                                                                                                                                   |
-| `table:use-regular-expressions` | formula matching rule — no formula engine                                                                                                                                                            |
-| `table:use-wildcards`           | formula matching rule — no formula engine                                                                                                                                                            |
+| name                            | why                                                                                                                  |
+| ------------------------------- | -------------------------------------------------------------------------------------------------------------------- |
+| `number:boolean-style`          | a data style for booleans. Excel has no format code for one — it shows TRUE/FALSE — so there is nothing to map it to |
+| `number:fill-character`         | the character Excel's `*` repeats to pad a cell. `*` is not in hucre's format model at either end                    |
+| `table:automatic-find-labels`   | formula label lookup — no formula engine                                                                             |
+| `table:calculation-settings`    | calculation preferences — hucre has no formula engine                                                                |
+| `table:case-sensitive`          | formula comparison rule — no formula engine                                                                          |
+| `table:iteration`               | iterative calculation settings — no formula engine                                                                   |
+| `table:maximum-difference`      | iteration convergence bound — no formula engine                                                                      |
+| `table:null-year`               | the century a two-digit year resolves into, for the formula parser                                                   |
+| `table:use-regular-expressions` | formula matching rule — no formula engine                                                                            |
+| `table:use-wildcards`           | formula matching rule — no formula engine                                                                            |
 
 ### Not named in the source
 
@@ -187,7 +186,7 @@ table:tracked-changes
 
 </details>
 
-<details><summary>164 attributes</summary>
+<details><summary>163 attributes</summary>
 
 ```
 number:calendar
@@ -246,7 +245,6 @@ table:database-table-name
 table:date-end
 table:date-start
 table:date-value
-table:default-cell-style-name
 table:display-border
 table:display-duplicates
 table:display-filter-buttons

scripts/spec-coverage.mjs

@@ -81,7 +81,7 @@
         const text = readFileSync(p, "utf8")
         corpusText += text + "\n"
         // A quoted literal — `case "sheetPr"` — is how a reader switches.
         for (const m of text.matchAll(/["'`]([A-Za-z_][\w:.\-]*)["'`]/g)) exact.add(m[1])
         // An object key — `xmlSelfClose("calcPr", { calcId: 0 })` — is how
         // a writer emits an attribute. Missing this form reported a dozen
         // attributes hucre writes on every workbook as unknown to it.
@@ -329,7 +329,7 @@
     "table:named-expressions":
       "read and written since #557, workbook-level. ODF scopes a name to a sheet by putting the block inside that `<table:table>`, which is not done yet; `<table:named-expression>` — a formula rather than a range — has no field to land in",
     "table:default-cell-style-name":
-      "**open** — a column's default cell style. hucre reads direct formatting only, and `PARITY.md` records that the ODS reader does not open `styles.xml`; this is the `content.xml` half of the same gap",
+      "read since #558 — a column's default fills in for cells that name no style of their own, which is where LibreOffice puts a column's number format. A column naming a style from `styles.xml` still resolves to nothing, and `PARITY.md` says so",
   }),
 )
 

src/ods/reader.ts

@@ -601,6 +601,30 @@ function parseCellValue(cell: XmlElement): CellValue {
  * `<table:named-expression>` — a formula rather than a range — has no
  * `NamedRange` to land in and is left alone.
  */
+/**
+ * The default cell style of each column, expanded by its repeat count.
+ *
+ * `table:default-cell-style-name` names an automatic style in
+ * `content.xml`, which this reader parses — so the format it points at is
+ * reachable. A column naming a style from `styles.xml` instead
+ * (LibreOffice writes `"Default"` on the columns it did not format)
+ * resolves to nothing, which is the right answer: that style *is* the
+ * absence of formatting, and `PARITY.md` records that `styles.xml` is not
+ * opened.
+ */
+function readColumnDefaultStyles(table: XmlElement): Array<string | undefined> {
+  const out: Array<string | undefined> = []
+  for (const column of findChildren(table, "table-column")) {
+    const style = column.attrs["table:default-cell-style-name"]
+    const repeat = Math.min(
+      Math.max(Number(column.attrs["table:number-columns-repeated"] ?? "1") || 1, 1),
+      MAX_COL_INDEX + 1,
+    )
+    for (let i = 0; i < repeat && out.length <= MAX_COL_INDEX; i++) out.push(style)
+  }
+  return out
+}
+
 function parseNamedExpressions(spreadsheet: XmlElement): NamedRange[] | undefined {
   const block = findChild(spreadsheet, "named-expressions")
   if (!block) return undefined
@@ -687,6 +711,13 @@ function parseContentXml(
     const cells = new Map<string, Cell>()
     const tableRows = findChildren(table, "table-row")
 
+    // LibreOffice puts a column's format on the column rather than on its
+    // cells: `<table:table-column table:default-cell-style-name="ce1"/>`,
+    // where `ce1` names the data style. A date column's `yyyy-mm-dd` lives
+    // there and nowhere else, so a document read without this came back
+    // with its values and none of its formats. See #464.
+    const columnDefaultStyles = readColumnDefaultStyles(table)
+
     let currentRow = 0
     let pendingEmptyRows = 0
 
@@ -715,6 +746,10 @@ function parseContentXml(
         hyperlink?: Hyperlink
       }> = []
 
+      // Which column the next entry starts at, so a cell that names no
+      // style can be given its column's default.
+      let colIndex = 0
+
       for (const child of tableRow.children) {
         if (typeof child === "string") continue
         const local = child.local || child.tag
@@ -724,7 +759,9 @@ function parseContentXml(
           const colSpan = Number(child.attrs["table:number-columns-spanned"] ?? "1")
           const rowSpan = Number(child.attrs["table:number-rows-spanned"] ?? "1")
           const value = parseCellValue(child)
-          const styleName = child.attrs["table:style-name"]
+          // A cell's own style wins; the column's default only fills in
+          // for cells that named none.
+          const styleName = child.attrs["table:style-name"] ?? columnDefaultStyles[colIndex]
           const formulaAttr = child.attrs["table:formula"]
           const formula = formulaAttr ? odsFormulaToExcel(formulaAttr) : undefined
           const { hyperlink } = extractTextAndHyperlink(child)
@@ -738,6 +775,7 @@ function parseContentXml(
             formula,
             hyperlink,
           })
+          colIndex += Number.isFinite(colRepeat) && colRepeat > 0 ? colRepeat : 1
         } else if (local === "covered-table-cell") {
           const colRepeat = Number(child.attrs["table:number-columns-repeated"] ?? "1")
           cellEntries.push({
@@ -747,6 +785,10 @@ function parseContentXml(
             rowSpan: 1,
             isCovered: true,
           })
+          // A covered cell still occupies its columns, so the count has
+          // to advance or every default after a merge lands one column
+          // to the left.
+          colIndex += Number.isFinite(colRepeat) && colRepeat > 0 ? colRepeat : 1
         }
       }
 

test/ods-column-default-style.test.ts

@@ -0,0 +1,89 @@
+import { describe, expect, it } from "vitest"
+import { readFileSync } from "node:fs"
+import { readOds } from "../src/ods/reader"
+import { writeOds } from "../src/ods/writer"
+
+// ═══════════════════════════════════════════════════════════════════════
+// LibreOffice puts a column's format on the column, not on its cells.
+// `<table:table-column table:default-cell-style-name="ce1"/>` names an
+// automatic style, and that style names the data style — so a date
+// column's `yyyy-mm-dd` lives there and nowhere else. The cells
+// themselves say `table:style-name="Default"`, a *named* style in
+// `styles.xml`, which this reader does not open.
+//
+// hucre read neither, so a LibreOffice document came back with its values
+// and none of its formats.
+//
+// The last open item in `SPEC-COVERAGE.md`: `table:default-cell-style-name`
+// was in the grammar, in `libreoffice-basic.ods`, and nowhere in `src/`.
+//
+// Only the `content.xml` half is reachable. A column pointing at a named
+// style — `"Default"`, which LibreOffice writes on the unformatted
+// columns — still resolves to nothing, because that style is in
+// `styles.xml`. PARITY records that, and this does not change it.
+// ═══════════════════════════════════════════════════════════════════════
+
+const LIBREOFFICE = "test/fixtures/third-party/libreoffice-basic.ods"
+
+describe("a column's default style reaches its cells", () => {
+  it("gives the LibreOffice date column its format back", async () => {
+    const bytes = new Uint8Array(readFileSync(LIBREOFFICE))
+    const wb = await readOds(bytes, { readStyles: true })
+    const sheet = wb.sheets[0]!
+
+    // Column C (index 2) is the date column; its style is on the column.
+    const dateCell = sheet.cells?.get("1,2")
+
+    expect(dateCell?.style?.numFmt).toBe("yyyy-mm-dd")
+  })
+
+  it("without readStyles, nothing changes", async () => {
+    // Styles are not being read, so a column default is not information
+    // the caller asked for — the same rule the rest of the reader follows.
+    const bytes = new Uint8Array(readFileSync(LIBREOFFICE))
+    const wb = await readOds(bytes)
+
+    expect(wb.sheets[0]!.cells?.get("1,2")?.style).toBeUndefined()
+  })
+
+  it("and the values are untouched either way", async () => {
+    const bytes = new Uint8Array(readFileSync(LIBREOFFICE))
+    const wb = await readOds(bytes, { readStyles: true })
+
+    expect(wb.sheets[0]!.rows[0]).toEqual(["Name", "Qty", "Date", "Active", "Total"])
+    expect(wb.sheets[0]!.rows[1]![2]).toBeInstanceOf(Date)
+  })
+})
+
+describe("a cell's own style still wins", () => {
+  it("over the column it sits in", async () => {
+    // hucre's own writer puts the style on the cell. That must not be
+    // overridden by a column default, whichever order they are read in.
+    const bytes = await writeOds({
+      sheets: [
+        {
+          name: "S",
+          rows: [[1, 2]],
+          cells: new Map([["0,0", { value: 1, style: { numFmt: "0.000" } }]]),
+        },
+      ],
+    })
+    const cell = (await readOds(bytes, { readStyles: true })).sheets[0]!.cells?.get("0,0")
+
+    expect(cell?.style?.numFmt).toBe("0.000")
+  })
+})
+
+describe("what is still not reachable", () => {
+  it("a column naming a style from styles.xml resolves to nothing", async () => {
+    // LibreOffice writes `table:default-cell-style-name="Default"` on the
+    // columns it did not format. `Default` lives in `styles.xml`, which
+    // this reader does not open, so those cells stay unstyled — which is
+    // the correct answer here, since the default style is the absence of
+    // formatting.
+    const bytes = new Uint8Array(readFileSync(LIBREOFFICE))
+    const wb = await readOds(bytes, { readStyles: true })
+
+    expect(wb.sheets[0]!.cells?.get("1,0")?.style?.numFmt).toBeUndefined()
+  })
+})

tsconfig.cli.json

@@ -29,6 +29,7 @@
     "test/xlsb-short-records.test.ts",
     "test/xlsx-stream-parity.test.ts",
     "test/indexed-color-palette.test.ts",
+    "test/ods-column-default-style.test.ts",
     "test/write-model.test.ts",
     "test/builder-coverage.test.ts"
   ],

tsconfig.json

@@ -49,6 +49,7 @@
     "test/xlsb-short-records.test.ts",
     "test/xlsx-stream-parity.test.ts",
     "test/indexed-color-palette.test.ts",
+    "test/ods-column-default-style.test.ts",
     "test/builder-coverage.test.ts"
   ]
 }