ferroxl 0.1.7

ferroxl: read and write Excel 2007 xlsx/xlsm files with an openpyxl-compatible API
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
//! Parts of a package ferroxl does not model, kept so a round trip does not lose them.
//!
//! # Why this is not just "copy the bytes"
//!
//! An enterprise workbook is mostly parts this library has never heard of: pivot tables and
//! their caches, slicers, query tables, connections, external links, threaded comments,
//! ActiveX controls, `vbaProject.bin`. Preserving those files is the easy half. The hard half
//! is that a part nobody can reach is inert, and in a package the three ways of reaching a part
//! have to travel together:
//!
//! 1. **The part itself.** Its bytes, written back unchanged.
//! 2. **Its content type.** `[Content_Types].xml` has to declare it, or Excel reports the
//!    package as corrupt before it looks at anything else.
//! 3. **A relationship to it.** `<pivotCache r:id="rId4"/>` in `xl/workbook.xml` resolves
//!    through `xl/_rels/workbook.xml.rels`.
//!
//! Drop (3) and keep (1) and you get inert data: a file that opens, shows the right cells, and
//! quietly has no pivot table. So the relationship is preserved too, and its id remapped when
//! the writer has already used that id for something of its own -- with the `r:id` reference in
//! the preserved element rewritten to match.
//!
//! # What still does not survive
//!
//! - **Unknown *attributes*.** Children are carried over; attributes on `<worksheet>`,
//!   `<sheetPr>` and the like are not. An attribute ferroxl does not understand is lost.
//! - **Anything outside a `<workbook>` or `<worksheet>` root child.** Shared formulas' master
//!   cells, for instance, live on `<c>` elements that are rebuilt from the cell model.
//! - **Byte-identical zip entries.** The preserved bytes are re-compressed, so an entry is
//!   content-identical rather than byte-identical. Reproducing the original deflate stream would
//!   tie the output to one compressor version.
//! - **Parts the writer emits.** Those are the writer's to produce, and a preserved copy would
//!   collide; `reader::preserved::is_writer_owned` decides which those are.

use crate::xml::functions::Element;
use std::collections::{BTreeMap, BTreeSet};

/// Children of `<workbook>` the writer does not emit.
const PRESERVED_WORKBOOK_CHILDREN: [&str; 11] = [
    "fileVersion",
    "functionGroups",
    "externalReferences",
    "oleSize",
    "customWorkbookViews",
    "pivotCaches",
    "smartTagPr",
    "smartTagTypes",
    "webPublishing",
    "fileRecoveryPr",
    "extLst",
];

/// Children of `<worksheet>` that may be preserved, because the writer does not produce them.
///
/// A superset, not the final answer. `<drawing>`, `<legacyDrawing>` and `<tableParts>` are in it
/// because a *loaded* sheet's are preserved -- ferroxl does not read charts or legacy drawings
/// back, so its writer produces none for a loaded workbook and the originals are all that is
/// left of them. The writer drops the ones it does produce for a sheet it is building, because a
/// duplicated `<drawing>` or `<tableParts>` is a file Excel reports as corrupt.
///
/// The elements genuinely absent from the writer are preserved unconditionally. Several are on
/// the parity roadmap; when one is implemented it comes out of this list in the same change, or
/// it will be written twice.
const CANDIDATE_WORKSHEET_CHILDREN: [&str; 23] = [
    "pivotTableParts",
    "extLst",
    "legacyDrawing",
    "legacyDrawingHF",
    "drawingHF",
    "picture",
    "oleObjects",
    "controls",
    "webPublishItems",
    "customSheetViews",
    "sheetProtection",
    "phoneticPr",
    "smartTags",
    "protectedRanges",
    "scenarios",
    "cellWatches",
    "dataConsolidate",
    "drawing",
    "tableParts",
    "ignoredErrors",
    "customProperties",
    "oleSize",
    "sheetCalcPr",
];

/// A relationship carried over from a loaded file.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct PreservedRelationship {
    /// The relationship id, as written in the source. May be remapped on the way out.
    pub id: String,
    /// The relationship type URI.
    pub relationship_type: String,
    /// The target, relative to the part holding the `.rels`.
    pub target: String,
    /// `TargetMode`, which is `External` for a link rather than an embedded part.
    pub target_mode: Option<String>,
}

impl PreservedRelationship {
    /// A relationship to a part inside the package.
    pub fn internal(
        id: impl Into<String>,
        relationship_type: impl Into<String>,
        target: impl Into<String>,
    ) -> Self {
        PreservedRelationship {
            id: id.into(),
            relationship_type: relationship_type.into(),
            target: target.into(),
            target_mode: None,
        }
    }

    /// A relationship to something outside the package.
    pub fn external(
        id: impl Into<String>,
        relationship_type: impl Into<String>,
        target: impl Into<String>,
    ) -> Self {
        PreservedRelationship {
            id: id.into(),
            relationship_type: relationship_type.into(),
            target: target.into(),
            target_mode: Some("External".to_string()),
        }
    }
}

/// A relationship the writer is producing itself.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct GeneratedRelationship {
    /// The id the writer assigned.
    pub id: String,
    /// The relationship type URI.
    pub relationship_type: String,
    /// The target, relative to the part holding the `.rels`.
    pub target: String,
    /// `TargetMode`, when the target is outside the package.
    pub target_mode: Option<String>,
}

impl GeneratedRelationship {
    /// A relationship to a part inside the package.
    pub fn internal(
        id: impl Into<String>,
        relationship_type: impl Into<String>,
        target: impl Into<String>,
    ) -> Self {
        GeneratedRelationship {
            id: id.into(),
            relationship_type: relationship_type.into(),
            target: target.into(),
            target_mode: None,
        }
    }

    /// This relationship as XML attributes, ready for a `<Relationship>` element.
    pub fn to_attributes(&self) -> Vec<(&'static str, String)> {
        let mut attributes = vec![
            ("Id", self.id.clone()),
            ("Type", self.relationship_type.clone()),
            ("Target", self.target.clone()),
        ];
        if let Some(mode) = &self.target_mode {
            attributes.push(("TargetMode", mode.clone()));
        }
        attributes
    }
}

/// The result of merging generated relationships with preserved ones.
#[derive(Debug, Default)]
pub struct MergedRelationships {
    /// Every relationship, in the order they should be written.
    pub relationships: Vec<GeneratedRelationship>,
    /// Preserved ids that had to move, from the source id to the id actually written.
    ///
    /// Needed because the elements referencing them are written too, and a
    /// `<pivotCache r:id="rId4"/>` pointing at a relationship that is now `rId9` resolves to
    /// nothing -- silently, which is the failure this whole module exists to avoid.
    pub id_map: BTreeMap<String, String>,
}

/// A `[Content_Types].xml` entry.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct PreservedContentType {
    /// `Default` (keyed by extension) or `Override` (keyed by part name).
    pub kind: String,
    /// The extension or the part name, depending on `kind`.
    pub name: String,
    /// The MIME type.
    pub content_type: String,
}

impl PreservedContentType {
    /// The attribute name this entry uses.
    ///
    /// The two spellings differ, and getting them the wrong way round produces a package Excel
    /// silently refuses.
    pub fn key_attribute(&self) -> &'static str {
        if self.kind == "Default" {
            "Extension"
        } else {
            "PartName"
        }
    }

    /// Whether this entry is an `Override` naming this exact part.
    pub fn covers_part(&self, path: &str) -> bool {
        self.kind != "Default" && self.name.trim_start_matches('/') == path
    }

    /// Whether this entry is a `Default` covering a part's extension.
    pub fn covers_extension(&self, extension: &str) -> bool {
        self.kind == "Default" && self.name.eq_ignore_ascii_case(extension)
    }
}

/// Everything a loaded workbook carried that the writer does not produce itself.
///
/// `Default` for an empty value, so a workbook built from scratch has none of it and pays
/// nothing.
#[derive(Debug, Clone, Default)]
pub struct PreservedParts {
    parts: BTreeMap<String, Vec<u8>>,
    content_types: Vec<PreservedContentType>,
    relationships: BTreeMap<String, Vec<PreservedRelationship>>,
    workbook_children: Vec<Element>,
    worksheet_children: BTreeMap<usize, Vec<Element>>,
    macro_extension: Option<String>,
    macro_enabled: bool,
}

impl PreservedParts {
    /// Whether anything at all was preserved.
    pub fn is_empty(&self) -> bool {
        self.parts.is_empty()
            && self.relationships.is_empty()
            && self.workbook_children.is_empty()
            && self.worksheet_children.is_empty()
    }

    /// Record a part's bytes.
    pub fn add_part(&mut self, path: impl Into<String>, data: Vec<u8>) {
        self.parts.insert(path.into(), data);
    }

    /// Record a content type entry.
    pub fn add_content_type(&mut self, entry: PreservedContentType) {
        self.content_types.push(entry);
    }

    /// Record the relationships held in a `.rels` part.
    pub fn add_relationships(
        &mut self,
        rels_path: impl Into<String>,
        relationships: Vec<PreservedRelationship>,
    ) {
        if !relationships.is_empty() {
            self.relationships.insert(rels_path.into(), relationships);
        }
    }

    /// Record a `<workbook>` child that the writer does not emit.
    pub fn add_workbook_child(&mut self, child: Element) {
        self.workbook_children.push(child);
    }

    /// Record the `<workbook>` children the writer does not emit.
    pub fn capture_workbook_children(&mut self, root: &Element) {
        self.workbook_children = root
            .children()
            .into_iter()
            .filter(|child| {
                local_name(&child.tag)
                    .is_some_and(|name| PRESERVED_WORKBOOK_CHILDREN.contains(&name))
            })
            .cloned()
            .collect();
    }

    /// Record the `<worksheet>` children that may need preserving, for a sheet.
    ///
    /// Which of these actually get written is the writer's call: it knows whether this sheet has
    /// charts, images, tables or comments, and so whether it will produce a `<drawing>`,
    /// `<legacyDrawing>` or `<tableParts>` of its own.
    pub fn capture_worksheet_children(&mut self, sheet: usize, root: &Element) {
        let children: Vec<Element> = root
            .children()
            .into_iter()
            .filter(|child| {
                local_name(&child.tag)
                    .is_some_and(|name| CANDIDATE_WORKSHEET_CHILDREN.contains(&name))
            })
            .cloned()
            .collect();
        if !children.is_empty() {
            self.worksheet_children.insert(sheet, children);
        }
    }

    /// Note that the source workbook was macro-enabled, and of what kind.
    pub fn set_macro_enabled(&mut self, extension: Option<String>) {
        self.macro_enabled = true;
        self.macro_extension = extension;
    }

    /// Whether the source workbook was macro-enabled.
    pub fn is_macro_enabled(&self) -> bool {
        self.macro_enabled
    }

    /// The source's macro-enabled extension, if it was one.
    pub fn macro_extension(&self) -> Option<&str> {
        self.macro_extension.as_deref()
    }

    /// The preserved part bytes, by path.
    pub fn parts(&self) -> &BTreeMap<String, Vec<u8>> {
        &self.parts
    }

    /// One preserved part's bytes.
    pub fn part(&self, path: &str) -> Option<&[u8]> {
        self.parts.get(path).map(|data| data.as_slice())
    }

    /// The preserved content type entries.
    pub fn content_types(&self) -> &[PreservedContentType] {
        &self.content_types
    }

    /// The preserved relationships held in one `.rels` part.
    pub fn relationships(&self, rels_path: &str) -> Option<&[PreservedRelationship]> {
        self.relationships
            .get(rels_path)
            .map(|list| list.as_slice())
    }

    /// Every `.rels` part that has preserved relationships, with them.
    pub fn all_relationships(&self) -> &BTreeMap<String, Vec<PreservedRelationship>> {
        &self.relationships
    }

    /// The preserved `<workbook>` children.
    pub fn workbook_children(&self) -> &[Element] {
        &self.workbook_children
    }

    /// The preserved `<worksheet>` children for a sheet, by zero-based index.
    pub fn worksheet_children(&self, sheet: usize) -> &[Element] {
        self.worksheet_children
            .get(&sheet)
            .map(|list| list.as_slice())
            .unwrap_or_default()
    }

    /// The preserved children for a sheet, minus any the writer is producing itself.
    ///
    /// `produced` names the child tags the writer has already emitted for this sheet. Without
    /// that argument a caller would have to know the writer's mind, and get it wrong by writing
    /// the same element twice.
    pub fn worksheet_children_excluding(&self, sheet: usize, produced: &[&str]) -> Vec<&Element> {
        self.worksheet_children(sheet)
            .iter()
            .filter(|child| local_name(&child.tag).is_none_or(|name| !produced.contains(&name)))
            .collect()
    }

    /// Whether this sheet index has preserved children worth writing.
    pub fn has_worksheet_children(&self, sheet: usize) -> bool {
        self.worksheet_children
            .get(&sheet)
            .is_some_and(|list| !list.is_empty())
    }

    /// The paths of preserved parts that no content type yet declares.
    ///
    /// A part with no declared type is a package Excel refuses, so this is the net for a source
    /// that was itself incomplete.
    pub fn untyped_part_paths(&self) -> Vec<&str> {
        self.parts
            .keys()
            .filter(|path| {
                let extension = path.rsplit('.').next().unwrap_or_default();
                !self
                    .content_types
                    .iter()
                    .any(|entry| entry.covers_part(path) || entry.covers_extension(extension))
            })
            .map(|path| path.as_str())
            .collect()
    }
}

/// The local part of a `{namespace}local` tag.
fn local_name(tag: &str) -> Option<&str> {
    tag.rsplit_once('}').map(|(_, local)| local)
}

/// Which relationship ids are already spoken for.
///
/// The writer numbers its own from `rId1` in every `.rels` part it writes, and a loaded
/// workbook's ids are usually also `rId1..rIdN` for sheets, styles and theme. So a preserved
/// `rId4` is very likely to collide, and emitting both would leave two parts claiming one id.
#[derive(Debug, Default)]
pub struct IdAllocator {
    used: BTreeSet<String>,
}

impl IdAllocator {
    /// Reserve the ids a writer has already emitted.
    pub fn reserve(&mut self, ids: impl IntoIterator<Item = String>) {
        self.used.extend(ids);
    }

    /// Map a source id to a free one, recording the choice.
    ///
    /// An id that is already free is kept, so a workbook with nothing to remap comes out with
    /// the ids it went in with -- which keeps `r:id` references in preserved elements valid
    /// without a rewrite.
    pub fn allocate(&mut self, wanted: &str) -> String {
        if !self.used.contains(wanted) {
            self.used.insert(wanted.to_string());
            return wanted.to_string();
        }
        // `rId` with a number is the overwhelmingly common form, so the search starts there
        // rather than at a hash that would produce an id no tool expects to see.
        let prefix = match wanted.strip_prefix("rId") {
            Some(rest) if !rest.is_empty() && rest.chars().all(|c| c.is_ascii_digit()) => {
                "rId".to_string()
            }
            _ => format!("{wanted}_"),
        };
        let mut counter = 1usize;
        loop {
            let candidate = format!("{prefix}{counter}");
            if !self.used.contains(&candidate) {
                self.used.insert(candidate.clone());
                return candidate;
            }
            counter += 1;
        }
    }

    /// Whether `id` is already taken.
    pub fn is_used(&self, id: &str) -> bool {
        self.used.contains(id)
    }
}

/// Merge the writer's relationships with a source's, moving any preserved id that would collide.
///
/// A preserved relationship whose target the writer also produces is dropped: two relationships
/// to one part is not what the source said, and keeping both would make Excel read whichever
/// target an element happened to name.
pub fn merge_relationships(
    generated: &[GeneratedRelationship],
    preserved: &[PreservedRelationship],
) -> MergedRelationships {
    let mut allocator = IdAllocator::default();
    allocator.reserve(generated.iter().map(|entry| entry.id.clone()));

    let generated_targets: BTreeSet<&str> = generated
        .iter()
        .map(|entry| entry.target.as_str())
        .collect();

    let mut merged = MergedRelationships {
        relationships: generated.to_vec(),
        id_map: BTreeMap::new(),
    };
    for entry in preserved {
        if entry.target_mode.is_none() && generated_targets.contains(entry.target.as_str()) {
            continue;
        }
        let id = allocator.allocate(&entry.id);
        if id != entry.id {
            merged.id_map.insert(entry.id.clone(), id.clone());
        }
        merged.relationships.push(GeneratedRelationship {
            id,
            relationship_type: entry.relationship_type.clone(),
            target: entry.target.clone(),
            target_mode: entry.target_mode.clone(),
        });
    }
    merged
}

/// Rewrite every `r:id` in an element subtree through `id_map`.
///
/// A no-op for an id that did not move, which is the common case.
pub fn rewrite_relationship_ids(element: &mut Element, id_map: &BTreeMap<String, String>) {
    if id_map.is_empty() {
        return;
    }
    for (key, value) in element.attributes.iter_mut() {
        let is_relationship_id = key == "r:id" || (key.starts_with('{') && key.ends_with("}id"));
        if is_relationship_id {
            if let Some(moved) = id_map.get(value.as_str()) {
                *value = moved.clone();
            }
        }
    }
    for child in element.children.iter_mut() {
        rewrite_relationship_ids(child, id_map);
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::xml::functions::fromstring;

    #[test]
    fn an_id_that_is_free_is_kept_so_references_stay_valid() {
        let mut allocator = IdAllocator::default();
        allocator.reserve(["rId1".to_string(), "rId2".to_string()]);
        assert_eq!(allocator.allocate("rId7"), "rId7");
    }

    #[test]
    fn a_colliding_id_is_moved_rather_than_duplicated() {
        let mut allocator = IdAllocator::default();
        allocator.reserve(["rId1".to_string(), "rId2".to_string()]);
        // The source's `rId2` pointed at a pivot cache. Emitting it as `rId2` too would leave two
        // parts claiming one id.
        assert_eq!(allocator.allocate("rId2"), "rId3");
        assert_eq!(allocator.allocate("rId3"), "rId4");
    }

    #[test]
    fn a_non_numeric_id_gets_a_suffixed_replacement() {
        let mut allocator = IdAllocator::default();
        allocator.reserve(["pkg1".to_string()]);
        assert_eq!(allocator.allocate("pkg1"), "pkg1_1");
    }

    #[test]
    fn nothing_is_reported_preserved_for_an_empty_value() {
        assert!(PreservedParts::default().is_empty());
    }

    #[test]
    fn a_part_without_a_content_type_is_reported() {
        let mut parts = PreservedParts::default();
        parts.add_part("xl/pivotTables/pivotTable1.xml", vec![1]);
        parts.add_part("xl/media/image9.png", vec![2]);
        parts.add_content_type(PreservedContentType {
            kind: "Default".to_string(),
            name: "png".to_string(),
            content_type: "image/png".to_string(),
        });
        assert_eq!(
            parts.untyped_part_paths(),
            vec!["xl/pivotTables/pivotTable1.xml"]
        );
    }

    #[test]
    fn an_override_satisfies_a_part_that_its_extension_would_not() {
        let mut parts = PreservedParts::default();
        parts.add_part("customXml/item1.xml", vec![1]);
        parts.add_content_type(PreservedContentType {
            kind: "Override".to_string(),
            name: "/customXml/item1.xml".to_string(),
            content_type: "application/xml".to_string(),
        });
        assert!(parts.untyped_part_paths().is_empty());
    }

    #[test]
    fn a_relationship_renders_the_two_attribute_spellings_correctly() {
        let embedded = GeneratedRelationship::internal("rId1", "type/a", "pivot1.xml");
        let attributes: BTreeMap<&str, String> = embedded.to_attributes().into_iter().collect();
        assert_eq!(
            attributes.get("Target").map(String::as_str),
            Some("pivot1.xml")
        );
        assert!(!attributes.contains_key("TargetMode"));

        let linked = GeneratedRelationship {
            target_mode: Some("External".to_string()),
            ..GeneratedRelationship::internal("rId2", "type/b", "https://example.com")
        };
        let attributes: BTreeMap<&str, String> = linked.to_attributes().into_iter().collect();
        assert_eq!(
            attributes.get("TargetMode").map(String::as_str),
            Some("External")
        );
    }

    #[test]
    fn a_content_type_reports_which_attribute_keys_it() {
        let by_extension = PreservedContentType {
            kind: "Default".to_string(),
            name: "bin".to_string(),
            content_type: "application/vnd.ms-office.vbaProject".to_string(),
        };
        let by_part = PreservedContentType {
            kind: "Override".to_string(),
            name: "/xl/workbook.xml".to_string(),
            content_type: "application/xml".to_string(),
        };
        assert_eq!(by_extension.key_attribute(), "Extension");
        assert_eq!(by_part.key_attribute(), "PartName");
    }

    #[test]
    fn a_colliding_preserved_id_is_remapped_and_the_reference_follows_it() {
        let generated = vec![
            GeneratedRelationship::internal("rId1", "t/sheet", "worksheets/sheet1.xml"),
            GeneratedRelationship::internal("rId2", "t/styles", "styles.xml"),
        ];
        let preserved = vec![PreservedRelationship::internal(
            "rId2",
            "t/pivotCacheDefinition",
            "pivotCache/pivotCacheDefinition1.xml",
        )];
        let merged = merge_relationships(&generated, &preserved);

        assert_eq!(merged.relationships.len(), 3);
        let moved = merged
            .relationships
            .last()
            .expect("the preserved one is last");
        assert_eq!(moved.id, "rId3", "it had to move: rId2 was taken");
        assert_eq!(
            merged.id_map.get("rId2").map(String::as_str),
            Some("rId3"),
            "and the reference has to be told where it went"
        );

        const REL_ID: &str =
            "{http://schemas.openxmlformats.org/officeDocument/2006/relationships}id";
        let mut element = Element::with_attributes("pivotCache", [(REL_ID, "rId2".to_string())]);
        rewrite_relationship_ids(&mut element, &merged.id_map);
        assert_eq!(element.get(REL_ID), Some("rId3"));
    }

    #[test]
    fn a_preserved_relationship_the_writer_also_produces_is_dropped() {
        // Two relationships to one part is not what the source said, and Excel would resolve
        // whichever id an element happened to name.
        let generated = vec![GeneratedRelationship::internal(
            "rId1",
            "t/theme",
            "theme/theme1.xml",
        )];
        let preserved = vec![PreservedRelationship::internal(
            "rId9",
            "t/theme",
            "theme/theme1.xml",
        )];
        let merged = merge_relationships(&generated, &preserved);
        assert_eq!(merged.relationships.len(), 1);
        assert!(
            merged.id_map.is_empty(),
            "nothing moved, so nothing needs rewriting"
        );
    }

    #[test]
    fn an_external_relationship_is_kept_even_when_the_target_looks_internal() {
        let generated = vec![GeneratedRelationship::internal(
            "rId1",
            "t/hyperlink",
            "https://example.com",
        )];
        let preserved = vec![PreservedRelationship::external(
            "rId5",
            "t/hyperlink",
            "https://example.com",
        )];
        let merged = merge_relationships(&generated, &preserved);
        assert_eq!(
            merged.relationships.len(),
            2,
            "a link is not the same as an embedded part"
        );
    }

    #[test]
    fn the_writer_produced_children_are_excluded_from_the_preserved_set() {
        let root = fromstring(
            br#"<worksheet xmlns="x">
                 <drawing r:id="rId1"/>
                 <pivotTableParts count="1"><pivotTablePart r:id="rId2"/></pivotTableParts>
               </worksheet>"#,
        )
        .expect("parses");
        let mut parts = PreservedParts::default();
        parts.capture_worksheet_children(0, &root);
        assert_eq!(parts.worksheet_children(0).len(), 2);
        assert_eq!(parts.worksheet_children_excluding(0, &["drawing"]).len(), 1);
    }

    #[test]
    fn a_workbook_child_the_writer_will_emit_is_not_preserved() {
        // `<sheets>` is written from the model. Preserving the source's would duplicate every
        // sheet.
        let root = fromstring(
            br#"<workbook xmlns="x"><sheets><sheet name="a"/></sheets><pivotCaches/></workbook>"#,
        )
        .expect("parses");
        let mut parts = PreservedParts::default();
        parts.capture_workbook_children(&root);
        assert_eq!(parts.workbook_children().len(), 1);
        // The tag keeps its namespace prefix, which is what makes re-emitting it valid; the
        // filtering is by local name because that is the part that identifies the element.
        assert!(
            parts.workbook_children()[0].tag.ends_with("pivotCaches"),
            "got {:?}",
            parts.workbook_children()[0].tag
        );
    }
}