max / pter

Initial implementation of pter (plain-text-email-renderer) HTML email to readable markdown converter. Handles element conversion, table layout unwrapping, reply chain detection, and tracking pixel stripping. 116 tests, criterion benchmarks, MIT licensed. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: Max J. <87768334+MaxJMath@users.noreply.github.com> · 2026-05-02 23:11 UTC

Commit: ef77e5f3ee2407db8d631c4200570f55c9342f54

17 files changed, +2890 insertions, -0 deletions

A .gitignore +1

			@@ -0,0 +1 @@
	1	+	/target

A Cargo.lock +500

		@@ -0,0 +1,1390 @@
1	+	# This file is automatically @generated by Cargo.
2	+	# It is not intended for manual editing.
3	+	version = 4
4	+
5	+	[[package]]
6	+	name = "aho-corasick"
7	+	version = "1.1.4"
8	+	source = "registry+https://github.com/rust-lang/crates.io-index"
9	+	checksum = "ddd31a130427c27518df266943a5308ed92d4b226cc639f5a8f1002816174301"
10	+	dependencies = [
11	+	"memchr",
12	+	]
13	+
14	+	[[package]]
15	+	name = "anes"
16	+	version = "0.1.6"
17	+	source = "registry+https://github.com/rust-lang/crates.io-index"
18	+	checksum = "4b46cbb362ab8752921c97e041f5e366ee6297bd428a31275b9fcf1e380f7299"
19	+
20	+	[[package]]
21	+	name = "anstyle"
22	+	version = "1.0.14"
23	+	source = "registry+https://github.com/rust-lang/crates.io-index"
24	+	checksum = "940b3a0ca603d1eade50a4846a2afffd5ef57a9feac2c0e2ec2e14f9ead76000"
25	+
26	+	[[package]]
27	+	name = "anyhow"
28	+	version = "1.0.102"
29	+	source = "registry+https://github.com/rust-lang/crates.io-index"
30	+	checksum = "7f202df86484c868dbad7eaa557ef785d5c66295e41b460ef922eca0723b842c"
31	+
32	+	[[package]]
33	+	name = "autocfg"
34	+	version = "1.5.0"
35	+	source = "registry+https://github.com/rust-lang/crates.io-index"
36	+	checksum = "c08606f8c3cbf4ce6ec8e28fb0014a2c086708fe954eaa885384a6165172e7e8"
37	+
38	+	[[package]]
39	+	name = "bit-set"
40	+	version = "0.8.0"
41	+	source = "registry+https://github.com/rust-lang/crates.io-index"
42	+	checksum = "08807e080ed7f9d5433fa9b275196cfc35414f66a0c79d864dc51a0d825231a3"
43	+	dependencies = [
44	+	"bit-vec",
45	+	]
46	+
47	+	[[package]]
48	+	name = "bit-vec"
49	+	version = "0.8.0"
50	+	source = "registry+https://github.com/rust-lang/crates.io-index"
51	+	checksum = "5e764a1d40d510daf35e07be9eb06e75770908c27d411ee6c92109c9840eaaf7"
52	+
53	+	[[package]]
54	+	name = "bitflags"
55	+	version = "2.11.1"
56	+	source = "registry+https://github.com/rust-lang/crates.io-index"
57	+	checksum = "c4512299f36f043ab09a583e57bceb5a5aab7a73db1805848e8fef3c9e8c78b3"
58	+
59	+	[[package]]
60	+	name = "bumpalo"
61	+	version = "3.20.2"
62	+	source = "registry+https://github.com/rust-lang/crates.io-index"
63	+	checksum = "5d20789868f4b01b2f2caec9f5c4e0213b41e3e5702a50157d699ae31ced2fcb"
64	+
65	+	[[package]]
66	+	name = "cast"
67	+	version = "0.3.0"
68	+	source = "registry+https://github.com/rust-lang/crates.io-index"
69	+	checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"
70	+
71	+	[[package]]
72	+	name = "cfg-if"
73	+	version = "1.0.4"
74	+	source = "registry+https://github.com/rust-lang/crates.io-index"
75	+	checksum = "9330f8b2ff13f34540b44e946ef35111825727b38d33286ef986142615121801"
76	+
77	+	[[package]]
78	+	name = "ciborium"
79	+	version = "0.2.2"
80	+	source = "registry+https://github.com/rust-lang/crates.io-index"
81	+	checksum = "42e69ffd6f0917f5c029256a24d0161db17cea3997d185db0d35926308770f0e"
82	+	dependencies = [
83	+	"ciborium-io",
84	+	"ciborium-ll",
85	+	"serde",
86	+	]
87	+
88	+	[[package]]
89	+	name = "ciborium-io"
90	+	version = "0.2.2"
91	+	source = "registry+https://github.com/rust-lang/crates.io-index"
92	+	checksum = "05afea1e0a06c9be33d539b876f1ce3692f4afea2cb41f740e7743225ed1c757"
93	+
94	+	[[package]]
95	+	name = "ciborium-ll"
96	+	version = "0.2.2"
97	+	source = "registry+https://github.com/rust-lang/crates.io-index"
98	+	checksum = "57663b653d948a338bfb3eeba9bb2fd5fcfaecb9e199e87e1eda4d9e8b240fd9"
99	+	dependencies = [
100	+	"ciborium-io",
101	+	"half",
102	+	]
103	+
104	+	[[package]]
105	+	name = "clap"
106	+	version = "4.6.1"
107	+	source = "registry+https://github.com/rust-lang/crates.io-index"
108	+	checksum = "1ddb117e43bbf7dacf0a4190fef4d345b9bad68dfc649cb349e7d17d28428e51"
109	+	dependencies = [
110	+	"clap_builder",
111	+	]
112	+
113	+	[[package]]
114	+	name = "clap_builder"
115	+	version = "4.6.0"
116	+	source = "registry+https://github.com/rust-lang/crates.io-index"
117	+	checksum = "714a53001bf66416adb0e2ef5ac857140e7dc3a0c48fb28b2f10762fc4b5069f"
118	+	dependencies = [
119	+	"anstyle",
120	+	"clap_lex",
121	+	]
122	+
123	+	[[package]]
124	+	name = "clap_lex"
125	+	version = "1.1.0"
126	+	source = "registry+https://github.com/rust-lang/crates.io-index"
127	+	checksum = "c8d4a3bb8b1e0c1050499d1815f5ab16d04f0959b233085fb31653fbfc9d98f9"
128	+
129	+	[[package]]
130	+	name = "criterion"
131	+	version = "0.5.1"
132	+	source = "registry+https://github.com/rust-lang/crates.io-index"
133	+	checksum = "f2b12d017a929603d80db1831cd3a24082f8137ce19c69e6447f54f5fc8d692f"
134	+	dependencies = [
135	+	"anes",
136	+	"cast",
137	+	"ciborium",
138	+	"clap",
139	+	"criterion-plot",
140	+	"is-terminal",
141	+	"itertools",
142	+	"num-traits",
143	+	"once_cell",
144	+	"oorandom",
145	+	"plotters",
146	+	"rayon",
147	+	"regex",
148	+	"serde",
149	+	"serde_derive",
150	+	"serde_json",
151	+	"tinytemplate",
152	+	"walkdir",
153	+	]
154	+
155	+	[[package]]
156	+	name = "criterion-plot"
157	+	version = "0.5.0"
158	+	source = "registry+https://github.com/rust-lang/crates.io-index"
159	+	checksum = "6b50826342786a51a89e2da3a28f1c32b06e387201bc2d19791f622c673706b1"
160	+	dependencies = [
161	+	"cast",
162	+	"itertools",
163	+	]
164	+
165	+	[[package]]
166	+	name = "crossbeam-deque"
167	+	version = "0.8.6"
168	+	source = "registry+https://github.com/rust-lang/crates.io-index"
169	+	checksum = "9dd111b7b7f7d55b72c0a6ae361660ee5853c9af73f70c3c2ef6858b950e2e51"
170	+	dependencies = [
171	+	"crossbeam-epoch",
172	+	"crossbeam-utils",
173	+	]
174	+
175	+	[[package]]
176	+	name = "crossbeam-epoch"
177	+	version = "0.9.18"
178	+	source = "registry+https://github.com/rust-lang/crates.io-index"
179	+	checksum = "5b82ac4a3c2ca9c3460964f020e1402edd5753411d7737aa39c3714ad1b5420e"
180	+	dependencies = [
181	+	"crossbeam-utils",
182	+	]
183	+
184	+	[[package]]
185	+	name = "crossbeam-utils"
186	+	version = "0.8.21"
187	+	source = "registry+https://github.com/rust-lang/crates.io-index"
188	+	checksum = "d0a5c400df2834b80a4c3327b3aad3a4c4cd4de0629063962b03235697506a28"
189	+
190	+	[[package]]
191	+	name = "crunchy"
192	+	version = "0.2.4"
193	+	source = "registry+https://github.com/rust-lang/crates.io-index"
194	+	checksum = "460fbee9c2c2f33933d720630a6a0bac33ba7053db5344fac858d4b8952d77d5"
195	+
196	+	[[package]]
197	+	name = "cssparser"
198	+	version = "0.36.0"
199	+	source = "registry+https://github.com/rust-lang/crates.io-index"
200	+	checksum = "dae61cf9c0abb83bd659dab65b7e4e38d8236824c85f0f804f173567bda257d2"
201	+	dependencies = [
202	+	"cssparser-macros",
203	+	"dtoa-short",
204	+	"itoa",
205	+	"phf",
206	+	"smallvec",
207	+	]
208	+
209	+	[[package]]
210	+	name = "cssparser-macros"
211	+	version = "0.6.1"
212	+	source = "registry+https://github.com/rust-lang/crates.io-index"
213	+	checksum = "13b588ba4ac1a99f7f2964d24b3d896ddc6bf847ee3855dbd4366f058cfcd331"
214	+	dependencies = [
215	+	"quote",
216	+	"syn",
217	+	]
218	+
219	+	[[package]]
220	+	name = "derive_more"
221	+	version = "2.1.1"
222	+	source = "registry+https://github.com/rust-lang/crates.io-index"
223	+	checksum = "d751e9e49156b02b44f9c1815bcb94b984cdcc4396ecc32521c739452808b134"
224	+	dependencies = [
225	+	"derive_more-impl",
226	+	]
227	+
228	+	[[package]]
229	+	name = "derive_more-impl"
230	+	version = "2.1.1"
231	+	source = "registry+https://github.com/rust-lang/crates.io-index"
232	+	checksum = "799a97264921d8623a957f6c3b9011f3b5492f557bbb7a5a19b7fa6d06ba8dcb"
233	+	dependencies = [
234	+	"proc-macro2",
235	+	"quote",
236	+	"rustc_version",
237	+	"syn",
238	+	]
239	+
240	+	[[package]]
241	+	name = "dtoa"
242	+	version = "1.0.11"
243	+	source = "registry+https://github.com/rust-lang/crates.io-index"
244	+	checksum = "4c3cf4824e2d5f025c7b531afcb2325364084a16806f6d47fbc1f5fbd9960590"
245	+
246	+	[[package]]
247	+	name = "dtoa-short"
248	+	version = "0.3.5"
249	+	source = "registry+https://github.com/rust-lang/crates.io-index"
250	+	checksum = "cd1511a7b6a56299bd043a9c167a6d2bfb37bf84a6dfceaba651168adfb43c87"
251	+	dependencies = [
252	+	"dtoa",
253	+	]
254	+
255	+	[[package]]
256	+	name = "ego-tree"
257	+	version = "0.11.0"
258	+	source = "registry+https://github.com/rust-lang/crates.io-index"
259	+	checksum = "b04dc5a38e4f151a79d9f2451ae6037fb6eaf5cba34771f44781f80e508498e3"
260	+
261	+	[[package]]
262	+	name = "either"
263	+	version = "1.15.0"
264	+	source = "registry+https://github.com/rust-lang/crates.io-index"
265	+	checksum = "48c757948c5ede0e46177b7add2e67155f70e33c07fea8284df6576da70b3719"
266	+
267	+	[[package]]
268	+	name = "equivalent"
269	+	version = "1.0.2"
270	+	source = "registry+https://github.com/rust-lang/crates.io-index"
271	+	checksum = "877a4ace8713b0bcf2a4e7eec82529c029f1d0619886d18145fea96c3ffe5c0f"
272	+
273	+	[[package]]
274	+	name = "errno"
275	+	version = "0.3.14"
276	+	source = "registry+https://github.com/rust-lang/crates.io-index"
277	+	checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
278	+	dependencies = [
279	+	"libc",
280	+	"windows-sys",
281	+	]
282	+
283	+	[[package]]
284	+	name = "fastrand"
285	+	version = "2.4.1"
286	+	source = "registry+https://github.com/rust-lang/crates.io-index"
287	+	checksum = "9f1f227452a390804cdb637b74a86990f2a7d7ba4b7d5693aac9b4dd6defd8d6"
288	+
289	+	[[package]]
290	+	name = "fnv"
291	+	version = "1.0.7"
292	+	source = "registry+https://github.com/rust-lang/crates.io-index"
293	+	checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
294	+
295	+	[[package]]
296	+	name = "foldhash"
297	+	version = "0.1.5"
298	+	source = "registry+https://github.com/rust-lang/crates.io-index"
299	+	checksum = "d9c4f5dac5e15c24eb999c26181a6ca40b39fe946cbe4c263c7209467bc83af2"
300	+
301	+	[[package]]
302	+	name = "futures-core"
303	+	version = "0.3.32"
304	+	source = "registry+https://github.com/rust-lang/crates.io-index"
305	+	checksum = "7e3450815272ef58cec6d564423f6e755e25379b217b0bc688e295ba24df6b1d"
306	+
307	+	[[package]]
308	+	name = "futures-task"
309	+	version = "0.3.32"
310	+	source = "registry+https://github.com/rust-lang/crates.io-index"
311	+	checksum = "037711b3d59c33004d3856fbdc83b99d4ff37a24768fa1be9ce3538a1cde4393"
312	+
313	+	[[package]]
314	+	name = "futures-util"
315	+	version = "0.3.32"
316	+	source = "registry+https://github.com/rust-lang/crates.io-index"
317	+	checksum = "389ca41296e6190b48053de0321d02a77f32f8a5d2461dd38762c0593805c6d6"
318	+	dependencies = [
319	+	"futures-core",
320	+	"futures-task",
321	+	"pin-project-lite",
322	+	"slab",
323	+	]
324	+
325	+	[[package]]
326	+	name = "getopts"
327	+	version = "0.2.24"
328	+	source = "registry+https://github.com/rust-lang/crates.io-index"
329	+	checksum = "cfe4fbac503b8d1f88e6676011885f34b7174f46e59956bba534ba83abded4df"
330	+	dependencies = [
331	+	"unicode-width",
332	+	]
333	+
334	+	[[package]]
335	+	name = "getrandom"
336	+	version = "0.3.4"
337	+	source = "registry+https://github.com/rust-lang/crates.io-index"
338	+	checksum = "899def5c37c4fd7b2664648c28120ecec138e4d395b459e5ca34f9cce2dd77fd"
339	+	dependencies = [
340	+	"cfg-if",
341	+	"libc",
342	+	"r-efi 5.3.0",
343	+	"wasip2",
344	+	]
345	+
346	+	[[package]]
347	+	name = "getrandom"
348	+	version = "0.4.2"
349	+	source = "registry+https://github.com/rust-lang/crates.io-index"
350	+	checksum = "0de51e6874e94e7bf76d726fc5d13ba782deca734ff60d5bb2fb2607c7406555"
351	+	dependencies = [
352	+	"cfg-if",
353	+	"libc",
354	+	"r-efi 6.0.0",
355	+	"wasip2",
356	+	"wasip3",
357	+	]
358	+
359	+	[[package]]
360	+	name = "half"
361	+	version = "2.7.1"
362	+	source = "registry+https://github.com/rust-lang/crates.io-index"
363	+	checksum = "6ea2d84b969582b4b1864a92dc5d27cd2b77b622a8d79306834f1be5ba20d84b"
364	+	dependencies = [
365	+	"cfg-if",
366	+	"crunchy",
367	+	"zerocopy",
368	+	]
369	+
370	+	[[package]]
371	+	name = "hashbrown"
372	+	version = "0.15.5"
373	+	source = "registry+https://github.com/rust-lang/crates.io-index"
374	+	checksum = "9229cfe53dfd69f0609a49f65461bd93001ea1ef889cd5529dd176593f5338a1"
375	+	dependencies = [
376	+	"foldhash",
377	+	]
378	+
379	+	[[package]]
380	+	name = "hashbrown"
381	+	version = "0.17.0"
382	+	source = "registry+https://github.com/rust-lang/crates.io-index"
383	+	checksum = "4f467dd6dccf739c208452f8014c75c18bb8301b050ad1cfb27153803edb0f51"
384	+
385	+	[[package]]
386	+	name = "heck"
387	+	version = "0.5.0"
388	+	source = "registry+https://github.com/rust-lang/crates.io-index"
389	+	checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
390	+
391	+	[[package]]
392	+	name = "hermit-abi"
393	+	version = "0.5.2"
394	+	source = "registry+https://github.com/rust-lang/crates.io-index"
395	+	checksum = "fc0fef456e4baa96da950455cd02c081ca953b141298e41db3fc7e36b1da849c"
396	+
397	+	[[package]]
398	+	name = "html5ever"
399	+	version = "0.39.0"
400	+	source = "registry+https://github.com/rust-lang/crates.io-index"
401	+	checksum = "46a1761807faccc9a19e86944bbf40610014066306f96edcdedc2fb714bcb7b8"
402	+	dependencies = [
403	+	"log",
404	+	"markup5ever",
405	+	]
406	+
407	+	[[package]]
408	+	name = "id-arena"
409	+	version = "2.3.0"
410	+	source = "registry+https://github.com/rust-lang/crates.io-index"
411	+	checksum = "3d3067d79b975e8844ca9eb072e16b31c3c1c36928edf9c6789548c524d0d954"
412	+
413	+	[[package]]
414	+	name = "indexmap"
415	+	version = "2.14.0"
416	+	source = "registry+https://github.com/rust-lang/crates.io-index"
417	+	checksum = "d466e9454f08e4a911e14806c24e16fba1b4c121d1ea474396f396069cf949d9"
418	+	dependencies = [
419	+	"equivalent",
420	+	"hashbrown 0.17.0",
421	+	"serde",
422	+	"serde_core",
423	+	]
424	+
425	+	[[package]]
426	+	name = "is-terminal"
427	+	version = "0.4.17"
428	+	source = "registry+https://github.com/rust-lang/crates.io-index"
429	+	checksum = "3640c1c38b8e4e43584d8df18be5fc6b0aa314ce6ebf51b53313d4306cca8e46"
430	+	dependencies = [
431	+	"hermit-abi",
432	+	"libc",
433	+	"windows-sys",
434	+	]
435	+
436	+	[[package]]
437	+	name = "itertools"
438	+	version = "0.10.5"
439	+	source = "registry+https://github.com/rust-lang/crates.io-index"
440	+	checksum = "b0fd2260e829bddf4cb6ea802289de2f86d6a7a690192fbe91b3f46e0f2c8473"
441	+	dependencies = [
442	+	"either",
443	+	]
444	+
445	+	[[package]]
446	+	name = "itoa"
447	+	version = "1.0.18"
448	+	source = "registry+https://github.com/rust-lang/crates.io-index"
449	+	checksum = "8f42a60cbdf9a97f5d2305f08a87dc4e09308d1276d28c869c684d7777685682"
450	+
451	+	[[package]]
452	+	name = "js-sys"
453	+	version = "0.3.97"
454	+	source = "registry+https://github.com/rust-lang/crates.io-index"
455	+	checksum = "a1840c94c045fbcf8ba2812c95db44499f7c64910a912551aaaa541decebcacf"
456	+	dependencies = [
457	+	"cfg-if",
458	+	"futures-util",
459	+	"once_cell",
460	+	"wasm-bindgen",
461	+	]
462	+
463	+	[[package]]
464	+	name = "leb128fmt"
465	+	version = "0.1.0"
466	+	source = "registry+https://github.com/rust-lang/crates.io-index"
467	+	checksum = "09edd9e8b54e49e587e4f6295a7d29c3ea94d469cb40ab8ca70b288248a81db2"
468	+
469	+	[[package]]
470	+	name = "libc"
471	+	version = "0.2.186"
472	+	source = "registry+https://github.com/rust-lang/crates.io-index"
473	+	checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66"
474	+
475	+	[[package]]
476	+	name = "linux-raw-sys"
477	+	version = "0.12.1"
478	+	source = "registry+https://github.com/rust-lang/crates.io-index"
479	+	checksum = "32a66949e030da00e8c7d4434b251670a91556f4144941d37452769c25d58a53"
480	+
481	+	[[package]]
482	+	name = "lock_api"
483	+	version = "0.4.14"
484	+	source = "registry+https://github.com/rust-lang/crates.io-index"
485	+	checksum = "224399e74b87b5f3557511d98dff8b14089b3dadafcab6bb93eab67d3aace965"
486	+	dependencies = [
487	+	"scopeguard",
488	+	]
489	+
490	+	[[package]]
491	+	name = "log"
492	+	version = "0.4.29"
493	+	source = "registry+https://github.com/rust-lang/crates.io-index"
494	+	checksum = "5e5032e24019045c762d3c0f28f5b6b8bbf38563a65908389bf7978758920897"
495	+
496	+	[[package]]
497	+	name = "markup5ever"
498	+	version = "0.39.0"
499	+	source = "registry+https://github.com/rust-lang/crates.io-index"
500	+	checksum = "7122d987ec5f704ee56f6e5b41a7d93722e9aae27ae07cafa4036c4d3f9757de"

Lines truncated

A Cargo.toml +20

		@@ -0,0 +1,20 @@
1	+	[package]
2	+	name = "pter"
3	+	version = "0.1.0"
4	+	edition = "2024"
5	+	description = "Plain Text Email Renderer — convert HTML email bodies into readable markdown"
6	+	license = "MIT"
7	+	repository = "https://github.com/maxjacobson/pter"
8	+	keywords = ["email", "html", "markdown", "plaintext", "converter"]
9	+	categories = ["email", "text-processing", "parser-implementations"]
10	+
11	+	[dependencies]
12	+	scraper = "0.26"
13	+
14	+	[dev-dependencies]
15	+	proptest = "1"
16	+	criterion = { version = "0.5", features = ["html_reports"] }
17	+
18	+	[[bench]]
19	+	name = "convert_bench"
20	+	harness = false

A LICENSE +21

		@@ -0,0 +1,21 @@
1	+	MIT License
2	+
3	+	Copyright (c) 2026 Max Jacobson
4	+
5	+	Permission is hereby granted, free of charge, to any person obtaining a copy
6	+	of this software and associated documentation files (the "Software"), to deal
7	+	in the Software without restriction, including without limitation the rights
8	+	to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9	+	copies of the Software, and to permit persons to whom the Software is
10	+	furnished to do so, subject to the following conditions:
11	+
12	+	The above copyright notice and this permission notice shall be included in all
13	+	copies or substantial portions of the Software.
14	+
15	+	THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16	+	IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17	+	FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18	+	AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19	+	LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20	+	OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21	+	SOFTWARE.

A README.md +33

		@@ -0,0 +1,33 @@
1	+	# pter
2	+
3	+	Plain Text Email Renderer — convert HTML email bodies into readable markdown.
4	+
5	+	Email HTML is a hostile environment: table-based layouts from 2004, tracking pixels,
6	+	Outlook conditional comments, deeply nested reply chains with inconsistent quoting.
7	+	pter converts all of it into clean, readable markdown.
8	+
9	+	## Usage
10	+
11	+	```rust
12	+	let html = r#"<h1>Hello</h1><p>This is an <strong>email</strong>.</p>"#;
13	+	let markdown = pter::convert(html);
14	+	assert_eq!(markdown, "# Hello\n\nThis is an email.");
15	+	```
16	+
17	+	## What it does
18	+
19	+	- Converts HTML elements to markdown equivalents (headings, links, lists, emphasis, code, images)
20	+	- Unwraps table-based email layouts (single-cell tables become content, multi-column linearizes)
21	+	- Detects and normalizes reply chains into `>` quoted markdown
22	+	- Strips tracking pixels, invisible elements, and Outlook conditional comments
23	+	- Produces output that is readable as plain text and renderable by any markdown toolchain
24	+
25	+	## What it does not do
26	+
27	+	- Parse MIME email structure (use `mailparse` or `mail-parser` for that)
28	+	- Extract article content from marketing templates (compose with a separate extractor)
29	+	- Render markdown to a display format (use `pulldown-cmark`, `comrak`, etc.)
30	+
31	+	## License
32	+
33	+	MIT

A benches/convert_bench.rs +97

		@@ -0,0 +1,97 @@
1	+	use criterion::{Criterion, black_box, criterion_group, criterion_main};
2	+
3	+	fn simple_email() -> &'static str {
4	+	r#"<html><body>
5	+	<h1>Meeting Tomorrow</h1>
6	+	<p>Hi Max,</p>
7	+	<p>Just confirming our meeting tomorrow at <strong>2pm</strong>.
8	+	Please review the <a href="https://example.com/doc">document</a> beforehand.</p>
9	+	<p>Best,<br>Alice</p>
10	+	</body></html>"#
11	+	}
12	+
13	+	fn newsletter_email() -> &'static str {
14	+	r#"<html><body>
15	+	<table width="100%" cellpadding="0" cellspacing="0" role="presentation">
16	+	<tr><td align="center">
17	+	<table width="600" cellpadding="0" cellspacing="0">
18	+	<tr><td>
19	+	<h2>Weekly Digest</h2>
20	+	<p>Here are your updates:</p>
21	+	<ul>
22	+	<li>New feature: <strong>Dark mode</strong> is now available</li>
23	+	<li>Bug fix: Resolved <a href="https://example.com/issue/123">issue #123</a></li>
24	+	<li>Update: API v2 documentation published</li>
25	+	</ul>
26	+	<p>Thanks for reading!</p>
27	+	<hr>
28	+	<p><small>Unsubscribe: <a href="https://example.com/unsub">click here</a></small></p>
29	+	</td></tr>
30	+	</table>
31	+	</td></tr>
32	+	</table>
33	+	<img src="https://track.example.com/open.gif" width="1" height="1">
34	+	</body></html>"#
35	+	}
36	+
37	+	fn reply_chain() -> &'static str {
38	+	r#"<html><body>
39	+	<div dir="ltr"><p>Got it, thanks!</p></div>
40	+	<div class="gmail_quote">
41	+	<div class="gmail_attr">On Tue, Jan 6, Bob wrote:</div>
42	+	<blockquote class="gmail_quote">
43	+	<div dir="ltr"><p>Here's the update you requested.</p></div>
44	+	<div class="gmail_quote">
45	+	<div class="gmail_attr">On Mon, Jan 5, Alice wrote:</div>
46	+	<blockquote class="gmail_quote">
47	+	<div dir="ltr"><p>What's the status on the deployment?</p></div>
48	+	</blockquote>
49	+	</div>
50	+	</blockquote>
51	+	</div>
52	+	</body></html>"#
53	+	}
54	+
55	+	fn large_email() -> String {
56	+	let paragraph = "<p>Lorem ipsum dolor sit amet, consectetur adipiscing elit. \
57	+	Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. \
58	+	Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris.</p>";
59	+	let mut html = String::from("<html><body>");
60	+	for i in 0..100 {
61	+	html.push_str(&format!("<h3>Section {}</h3>", i));
62	+	html.push_str(paragraph);
63	+	}
64	+	html.push_str("</body></html>");
65	+	html
66	+	}
67	+
68	+	fn bench_simple(c: &mut Criterion) {
69	+	let html = simple_email();
70	+	c.bench_function("simple_email", \|b\| {
71	+	b.iter(\|\| pter::convert(black_box(html)))
72	+	});
73	+	}
74	+
75	+	fn bench_newsletter(c: &mut Criterion) {
76	+	let html = newsletter_email();
77	+	c.bench_function("newsletter_layout_tables", \|b\| {
78	+	b.iter(\|\| pter::convert(black_box(html)))
79	+	});
80	+	}
81	+
82	+	fn bench_reply_chain(c: &mut Criterion) {
83	+	let html = reply_chain();
84	+	c.bench_function("reply_chain_nested", \|b\| {
85	+	b.iter(\|\| pter::convert(black_box(html)))
86	+	});
87	+	}
88	+
89	+	fn bench_large(c: &mut Criterion) {
90	+	let html = large_email();
91	+	c.bench_function("large_100_sections", \|b\| {
92	+	b.iter(\|\| pter::convert(black_box(&html)))
93	+	});
94	+	}
95	+
96	+	criterion_group!(benches, bench_simple, bench_newsletter, bench_reply_chain, bench_large);
97	+	criterion_main!(benches);

A docs/architecture.md +47

		@@ -0,0 +1,47 @@
1	+	# pter Architecture
2	+
3	+	## Overview
4	+
5	+	pter converts HTML email bodies into readable markdown. It takes an HTML string and returns a markdown string. It does not handle MIME parsing, content extraction, or markdown rendering.
6	+
7	+	## Pipeline
8	+
9	+	```
10	+	html: &str
11	+	→ scraper::Html::parse_document() # html5ever DOM tree
12	+	→ walk_children(root) # depth-first traversal
13	+	→ handle_text() # whitespace collapsing, entity decoding
14	+	→ handle_element() # classify → skip / transparent / block / inline
15	+	→ handle_block() # paragraphs, headings, lists, blockquotes, pre, hr
16	+	→ handle_inline() # bold, italic, links, images, code, br
17	+	→ whitespace::normalize() # collapse blank lines, trim
18	+	→ String
19	+	```
20	+
21	+	## Module Responsibilities
22	+
23	+	\| Module \| Responsibility \|
24	+	\|--------\|---------------\|
25	+	\| `lib.rs` \| Public API (`convert`), re-exports \|
26	+	\| `convert.rs` \| DOM walker, `Context` state, element dispatch \|
27	+	\| `elements.rs` \| Element classification, tracking pixel / hidden detection \|
28	+	\| `whitespace.rs` \| Output normalization \|
29	+	\| `tables.rs` \| Table layout detection and unwrapping (Phase 2) \|
30	+	\| `replies.rs` \| Reply chain detection and quoting (Phase 3) \|
31	+
32	+	## Design Decisions
33	+
34	+	scraper over html5ever directly: We need tree traversal (parent/child/sibling access) for layout table unwrapping and reply chain detection. scraper provides this via ego-tree on top of html5ever's spec-compliant parsing.
35	+
36	+	Markdown output: Markdown is readable as plain text and renderable by any toolchain. It preserves structural information (headings, links, lists) that plain text loses.
37	+
38	+	Faithful conversion: pter converts what's there. Content extraction (stripping marketing wrappers) and post-processing (trimming signatures) are separate concerns, composable before or after pter.
39	+
40	+	Blockquote rendering: Blockquotes render children into a temporary buffer, then prefix each line with `> `. This handles nested blockquotes naturally — inner quotes produce `> ` lines, outer quote prefixes them again to get `> > `.
41	+
42	+	## Dependencies
43	+
44	+	\| Crate \| Purpose \|
45	+	\|-------\|---------\|
46	+	\| `scraper` \| HTML parsing + DOM tree + CSS selectors \|
47	+	\| `proptest` (dev) \| Property-based testing \|

A docs/todo.md +95

		@@ -0,0 +1,95 @@
1	+	# pter - Todo
2	+
3	+	Done: Phases 1-5 (except publish). Active: None. Next: cargo publish when ready.
4	+
5	+	v0.1.0. 116 tests.
6	+
7	+	---
8	+
9	+	## Phase 1: Core Conversion
10	+
11	+	### Done
12	+	- [x] Crate scaffold (Cargo.toml, MIT license, README)
13	+	- [x] HTML element to markdown conversion (p, h1-h6, strong, em, a, img, ul/ol/li, blockquote, pre/code, hr, br, del, sup, sub)
14	+	- [x] Tracking pixel detection (1x1 img, empty src, data URI, inline style)
15	+	- [x] Hidden element skipping (display:none, visibility:hidden)
16	+	- [x] Whitespace normalization (collapse blank lines, trim)
17	+	- [x] Script/style/head stripping
18	+	- [x] Entity decoding (via html5ever)
19	+	- [x] Link deduplication (text matches URL)
20	+	- [x] Nested list indentation
21	+	- [x] Nested blockquote rendering
22	+	- [x] Pre/code block rendering (no double-wrap)
23	+
24	+	---
25	+
26	+	## Phase 2: Email Layout Unwrapping
27	+
28	+	### Done
29	+	- [x] Layout table detection heuristic (layout vs data table)
30	+	- [x] Single-cell table unwrapping
31	+	- [x] Multi-column table linearization
32	+	- [x] Data table rendering as markdown table
33	+	- [x] Nested layout table recursion
34	+	- [x] font-size:0 / line-height:0 / height:0+overflow:hidden spacer detection
35	+	- [x] role="presentation" detection
36	+
37	+	### Deferred
38	+	- [ ] Outlook conditional comment stripping (client-specific, low cross-platform value)
39	+
40	+	---
41	+
42	+	## Phase 3: Reply Chain Detection
43	+
44	+	### Done
45	+	- [x] Reply boundary abstraction (`is_reply_boundary` predicate)
46	+	- [x] Structural markers (type=cite)
47	+	- [x] CSS class markers (gmail_quote, divRplyFwdMsg, yahoo_quoted, protonmail_quote, tutanota_quote, moz-cite-prefix, zmail_extra)
48	+	- [x] Attribution text detection (On ... wrote:, Forwarded message, Original Message, Begin forwarded message, French/German variants)
49	+	- [x] Attribution line preservation above quote blocks
50	+	- [x] Quote depth rendering via temporary buffer + `>` prefix
51	+	- [x] Outlook separator detection (From/Sent/To/Subject blocks)
52	+	- [x] Heuristic: div with attribution text followed by blockquote
53	+	- [x] Previous sibling text scanning for attribution
54	+
55	+	---
56	+
57	+	## Phase 4: Integration
58	+
59	+	### Done
60	+	- [x] GoingsOn: pter::convert() replaces strip_html in imap_client.rs extract_body_with_html()
61	+	- [x] GoingsOn: removed ~230 lines of hand-rolled HTML stripping code + 30 tests (covered by pter)
62	+	- [x] GoingsOn: path dep added to src-tauri/Cargo.toml
63	+	- [x] Balanced Breakfast: pter::convert() replaces html2text in html_to_text + extract_article Rhai host functions
64	+	- [x] Balanced Breakfast: html2text dependency removed from bb-core/Cargo.toml
65	+	- [x] Both projects compile clean, BB tests pass (153 tests)
66	+
67	+	---
68	+
69	+	## Phase 5: Polish + Publish
70	+
71	+	### Done
72	+	- [x] Property-based testing with proptest (7 fuzz strategies: never panics, no HTML leak, valid UTF-8, no triple newlines, no trailing whitespace, arbitrary bytes, whitespace-only)
73	+	- [x] Edge case hardening (24 tests: empty, whitespace-only, deeply nested divs/blockquotes/lists, malformed HTML, unicode, large input, empty table cells, nested link formatting)
74	+	- [x] Benchmarks with criterion (simple: 4µs, newsletter: 15µs, reply chain: 10µs, 100 sections: 101µs)
75	+
76	+	### Remaining
77	+	- [ ] cargo publish to crates.io
78	+	- [ ] Update GO and BB to crates.io version
79	+
80	+	---
81	+
82	+	## Key Paths
83	+
84	+	\| What \| Where \|
85	+	\|------\|-------\|
86	+	\| Public API \| `src/lib.rs` \|
87	+	\| Conversion pipeline \| `src/convert.rs` \|
88	+	\| Element classification \| `src/elements.rs` \|
89	+	\| Table handling \| `src/tables.rs` \|
90	+	\| Reply detection \| `src/replies.rs` \|
91	+	\| Whitespace normalization \| `src/whitespace.rs` \|
92	+	\| Integration tests \| `tests/integration.rs` \|
93	+	\| Edge case tests \| `tests/edge_cases.rs` \|
94	+	\| Property-based tests \| `tests/proptest.rs` \|
95	+	\| Benchmarks \| `benches/convert_bench.rs` \|

A src/convert.rs +500

		@@ -0,0 +1,689 @@
1	+	use scraper::node::Node;
2	+	use scraper::{ElementRef, Html};
3	+
4	+	use crate::elements::{self, BlockKind, ElementAction, InlineKind};
5	+	use crate::replies;
6	+	use crate::tables;
7	+	use crate::whitespace;
8	+
9	+	/// Convert an HTML email body into readable markdown.
10	+	///
11	+	/// This is the main entry point for pter. Pass in an HTML string
12	+	/// (just the body, not MIME structure) and get back clean markdown.
13	+	///
14	+	/// ```
15	+	/// let md = pter::convert("<p>Hello <strong>world</strong></p>");
16	+	/// assert_eq!(md, "Hello world");
17	+	/// ```
18	+	pub fn convert(html: &str) -> String {
19	+	if html.is_empty() {
20	+	return String::new();
21	+	}
22	+
23	+	let document = Html::parse_document(html);
24	+	let mut ctx = Context::new();
25	+	walk_children(document.root_element(), &mut ctx);
26	+	whitespace::normalize(&ctx.output)
27	+	}
28	+
29	+	/// Conversion state threaded through the tree walk.
30	+	struct Context {
31	+	output: String,
32	+	/// Current list nesting depth (for indentation).
33	+	list_depth: u32,
34	+	/// Whether we're inside a <pre> block (preserve whitespace).
35	+	in_pre: bool,
36	+	/// Whether we're inside an <a> tag (don't nest links).
37	+	in_link: bool,
38	+	/// Stack of list types for proper ordered/unordered rendering.
39	+	list_stack: Vec<ListType>,
40	+	}
41	+
42	+	#[derive(Clone, Copy)]
43	+	enum ListType {
44	+	Unordered,
45	+	Ordered(u32), // current item number
46	+	}
47	+
48	+	impl Context {
49	+	fn new() -> Self {
50	+	Self {
51	+	output: String::with_capacity(4096),
52	+	list_depth: 0,
53	+	in_pre: false,
54	+	in_link: false,
55	+	list_stack: Vec::new(),
56	+	}
57	+	}
58	+
59	+	fn push(&mut self, s: &str) {
60	+	self.output.push_str(s);
61	+	}
62	+
63	+	fn push_char(&mut self, c: char) {
64	+	self.output.push(c);
65	+	}
66	+
67	+	fn ensure_blank_line(&mut self) {
68	+	let trimmed = self.output.trim_end_matches(' ');
69	+	if trimmed.is_empty() {
70	+	return;
71	+	}
72	+	if trimmed.ends_with("\n\n") {
73	+	return;
74	+	}
75	+	self.output.truncate(trimmed.len());
76	+	self.output.push_str("\n\n");
77	+	}
78	+
79	+	fn ensure_newline(&mut self) {
80	+	if !self.output.is_empty() && !self.output.ends_with('\n') {
81	+	self.output.push('\n');
82	+	}
83	+	}
84	+
85	+	fn list_indent(&self) -> String {
86	+	if self.list_depth <= 1 {
87	+	return String::new();
88	+	}
89	+	" ".repeat((self.list_depth - 1) as usize)
90	+	}
91	+	}
92	+
93	+	/// Walk all children of a node, converting each to markdown.
94	+	fn walk_children(parent: ElementRef, ctx: &mut Context) {
95	+	for child in parent.children() {
96	+	match child.value() {
97	+	Node::Text(text) => {
98	+	handle_text(&text.text, ctx);
99	+	}
100	+	Node::Element(_) => {
101	+	if let Some(el_ref) = ElementRef::wrap(child) {
102	+	handle_element(el_ref, ctx);
103	+	}
104	+	}
105	+	_ => {}
106	+	}
107	+	}
108	+	}
109	+
110	+	/// Handle a text node.
111	+	fn handle_text(text: &str, ctx: &mut Context) {
112	+	if ctx.in_pre {
113	+	ctx.push(text);
114	+	return;
115	+	}
116	+
117	+	// Collapse whitespace in normal flow
118	+	let mut last_was_space = ctx.output.ends_with(' ') \|\| ctx.output.ends_with('\n');
119	+	for ch in text.chars() {
120	+	if ch.is_ascii_whitespace() {
121	+	if !last_was_space {
122	+	ctx.push_char(' ');
123	+	last_was_space = true;
124	+	}
125	+	} else {
126	+	ctx.push_char(ch);
127	+	last_was_space = false;
128	+	}
129	+	}
130	+	}
131	+
132	+	/// Handle an element node — classify it and render accordingly.
133	+	fn handle_element(el: ElementRef, ctx: &mut Context) {
134	+	let element = el.value();
135	+
136	+	// Check hidden elements
137	+	if elements::is_hidden(element) {
138	+	return;
139	+	}
140	+
141	+	// Check for reply boundaries before normal classification.
142	+	// Reply boundaries (gmail_quote, type=cite, etc.) get rendered
143	+	// as blockquotes regardless of their actual element type.
144	+	if replies::is_reply_boundary(el) {
145	+	render_reply_block(el, ctx);
146	+	return;
147	+	}
148	+
149	+	// Check for Outlook-style "From: ... Sent: ..." separator blocks.
150	+	// These introduce quoted content that follows them.
151	+	if replies::is_outlook_separator(el) {
152	+	ctx.ensure_blank_line();
153	+	// Render the separator header as attribution
154	+	let text: String = el.text().collect();
155	+	let trimmed = text.split_whitespace().collect::<Vec<_>>().join(" ");
156	+	ctx.push(&trimmed);
157	+	ctx.ensure_blank_line();
158	+	return;
159	+	}
160	+
161	+	match elements::classify(element) {
162	+	ElementAction::Skip => {}
163	+	ElementAction::Transparent => walk_children(el, ctx),
164	+	ElementAction::Block(kind) => handle_block(el, ctx, kind),
165	+	ElementAction::Inline(kind) => handle_inline(el, ctx, kind),
166	+	}
167	+	}
168	+
169	+	fn handle_block(el: ElementRef, ctx: &mut Context, kind: BlockKind) {
170	+	match kind {
171	+	BlockKind::Paragraph => {
172	+	ctx.ensure_blank_line();
173	+	walk_children(el, ctx);
174	+	ctx.ensure_blank_line();
175	+	}
176	+
177	+	BlockKind::Heading(level) => {
178	+	ctx.ensure_blank_line();
179	+	let prefix = "#".repeat(level as usize);
180	+	ctx.push(&prefix);
181	+	ctx.push_char(' ');
182	+	walk_children(el, ctx);
183	+	ctx.ensure_blank_line();
184	+	}
185	+
186	+	BlockKind::Blockquote => {
187	+	ctx.ensure_blank_line();
188	+	// Render children into a temporary buffer, then prefix each line with >
189	+	let mut inner_ctx = Context::new();
190	+	inner_ctx.in_pre = ctx.in_pre;
191	+	inner_ctx.in_link = ctx.in_link;
192	+	walk_children(el, &mut inner_ctx);
193	+	let inner = whitespace::normalize(&inner_ctx.output);
194	+	for line in inner.lines() {
195	+	ctx.push("> ");
196	+	ctx.push(line);
197	+	ctx.push_char('\n');
198	+	}
199	+	ctx.push_char('\n');
200	+	}
201	+
202	+	BlockKind::UnorderedList => {
203	+	ctx.ensure_blank_line();
204	+	ctx.list_depth += 1;
205	+	ctx.list_stack.push(ListType::Unordered);
206	+	walk_children(el, ctx);
207	+	ctx.list_stack.pop();
208	+	ctx.list_depth -= 1;
209	+	ctx.ensure_blank_line();
210	+	}
211	+
212	+	BlockKind::OrderedList => {
213	+	ctx.ensure_blank_line();
214	+	ctx.list_depth += 1;
215	+	ctx.list_stack.push(ListType::Ordered(0));
216	+	walk_children(el, ctx);
217	+	ctx.list_stack.pop();
218	+	ctx.list_depth -= 1;
219	+	ctx.ensure_blank_line();
220	+	}
221	+
222	+	BlockKind::ListItem => {
223	+	ctx.ensure_newline();
224	+	let indent = ctx.list_indent();
225	+	ctx.push(&indent);
226	+
227	+	// Determine bullet or number
228	+	let marker = match ctx.list_stack.last_mut() {
229	+	Some(ListType::Unordered) => "- ".to_string(),
230	+	Some(ListType::Ordered(n)) => {
231	+	*n += 1;
232	+	format!("{}. ", *n)
233	+	}
234	+	None => "- ".to_string(),
235	+	};
236	+	ctx.push(&marker);
237	+	walk_children(el, ctx);
238	+	ctx.ensure_newline();
239	+	}
240	+
241	+	BlockKind::PreFormatted => {
242	+	ctx.ensure_blank_line();
243	+	ctx.push("```\n");
244	+	ctx.in_pre = true;
245	+	walk_children(el, ctx);
246	+	ctx.in_pre = false;
247	+	ctx.ensure_newline();
248	+	ctx.push("```");
249	+	ctx.ensure_blank_line();
250	+	}
251	+
252	+	BlockKind::HorizontalRule => {
253	+	ctx.ensure_blank_line();
254	+	ctx.push("---");
255	+	ctx.ensure_blank_line();
256	+	}
257	+
258	+	BlockKind::Table => {
259	+	ctx.ensure_blank_line();
260	+	if tables::is_data_table(el) {
261	+	let (headers, rows) = tables::extract_table_data(el);
262	+	let md = tables::render_markdown_table(&headers, &rows);
263	+	if !md.is_empty() {
264	+	ctx.push(&md);
265	+	}
266	+	} else {
267	+	// Layout table — unwrap and render cell contents directly
268	+	render_layout_table(el, ctx);
269	+	}
270	+	ctx.ensure_blank_line();
271	+	}
272	+
273	+	BlockKind::Div => {
274	+	// Divs act as block separators but don't add their own markup
275	+	ctx.ensure_blank_line();
276	+	walk_children(el, ctx);
277	+	ctx.ensure_blank_line();
278	+	}
279	+	}
280	+	}
281	+
282	+	fn handle_inline(el: ElementRef, ctx: &mut Context, kind: InlineKind) {
283	+	match kind {
284	+	InlineKind::Bold => {
285	+	ctx.push("**");
286	+	walk_children(el, ctx);
287	+	ctx.push("**");
288	+	}
289	+
290	+	InlineKind::Italic => {
291	+	ctx.push("*");
292	+	walk_children(el, ctx);
293	+	ctx.push("*");
294	+	}
295	+
296	+	InlineKind::Strikethrough => {
297	+	ctx.push("~~");
298	+	walk_children(el, ctx);
299	+	ctx.push("~~");
300	+	}
301	+
302	+	InlineKind::Code => {
303	+	if ctx.in_pre {
304	+	// Inside a <pre>, don't double-wrap
305	+	walk_children(el, ctx);
306	+	} else {
307	+	ctx.push("`");
308	+	walk_children(el, ctx);
309	+	ctx.push("`");
310	+	}
311	+	}
312	+
313	+	InlineKind::Link => {
314	+	if ctx.in_link {
315	+	// Don't nest links
316	+	walk_children(el, ctx);
317	+	return;
318	+	}
319	+
320	+	let href = el.value().attr("href").unwrap_or("");
321	+
322	+	if href.is_empty() \|\| href == "#" {
323	+	walk_children(el, ctx);
324	+	return;
325	+	}
326	+
327	+	// Collect the link text
328	+	let mut text_ctx = Context::new();
329	+	text_ctx.in_link = true;
330	+	walk_children(el, &mut text_ctx);
331	+	let text = text_ctx.output.trim().to_string();
332	+
333	+	if text.is_empty() {
334	+	// Link with no text — just show the URL
335	+	ctx.push(href);
336	+	} else if text == href {
337	+	// Link text matches URL — no need for markdown link syntax
338	+	ctx.push(href);
339	+	} else {
340	+	ctx.push("[");
341	+	ctx.push(&text);
342	+	ctx.push("](");
343	+	ctx.push(href);
344	+	ctx.push(")");
345	+	}
346	+	}
347	+
348	+	InlineKind::Image => {
349	+	let element = el.value();
350	+	if elements::is_tracking_pixel(element) {
351	+	return;
352	+	}
353	+
354	+	let alt = element.attr("alt").unwrap_or("");
355	+	let src = element.attr("src").unwrap_or("");
356	+
357	+	if src.is_empty() {
358	+	return;
359	+	}
360	+
361	+	ctx.push("![");
362	+	ctx.push(alt);
363	+	ctx.push("](");
364	+	ctx.push(src);
365	+	ctx.push(")");
366	+	}
367	+
368	+	InlineKind::LineBreak => {
369	+	ctx.push_char('\n');
370	+	}
371	+
372	+	InlineKind::Superscript => {
373	+	ctx.push("^");
374	+	walk_children(el, ctx);
375	+	}
376	+
377	+	InlineKind::Subscript => {
378	+	ctx.push("~");
379	+	walk_children(el, ctx);
380	+	}
381	+	}
382	+	}
383	+
384	+	/// Render a reply boundary as a quoted block.
385	+	///
386	+	/// This is the same rendering logic as `<blockquote>` — children are
387	+	/// rendered into a temporary buffer and each line gets `> ` prefixed.
388	+	/// Attribution lines (e.g. "On ... wrote:") are rendered above the quote.
389	+	fn render_reply_block(el: ElementRef, ctx: &mut Context) {
390	+	ctx.ensure_blank_line();
391	+
392	+	// Look for attribution text
393	+	if let Some(attribution) = replies::find_attribution(el) {
394	+	ctx.push(&attribution);
395	+	ctx.push_char('\n');
396	+	}
397	+
398	+	// Render children into temp buffer, then prefix with >
399	+	let mut inner_ctx = Context::new();
400	+	inner_ctx.in_pre = ctx.in_pre;
401	+	inner_ctx.in_link = ctx.in_link;
402	+	walk_children(el, &mut inner_ctx);
403	+	let inner = whitespace::normalize(&inner_ctx.output);
404	+
405	+	if !inner.is_empty() {
406	+	for line in inner.lines() {
407	+	ctx.push("> ");
408	+	ctx.push(line);
409	+	ctx.push_char('\n');
410	+	}
411	+	ctx.push_char('\n');
412	+	}
413	+	}
414	+
415	+	/// Unwrap a layout table by rendering cell contents sequentially.
416	+	///
417	+	/// Walks through rows and cells, rendering each cell's content as if
418	+	/// the table wrapper didn't exist. This handles the common email pattern
419	+	/// of wrapping everything in `<table><tr><td>...</td></tr></table>`.
420	+	fn render_layout_table(table: ElementRef, ctx: &mut Context) {
421	+	for descendant in table.descendants() {
422	+	if let Some(el_ref) = ElementRef::wrap(descendant) {
423	+	let name = el_ref.value().name();
424	+	if name == "td" \|\| name == "th" {
425	+	// Check if the cell itself is hidden
426	+	if !elements::is_hidden(el_ref.value()) {
427	+	walk_children(el_ref, ctx);
428	+	ctx.ensure_blank_line();
429	+	}
430	+	}
431	+	}
432	+	}
433	+	}
434	+
435	+	#[cfg(test)]
436	+	mod tests {
437	+	use super::*;
438	+
439	+	// -- Basic elements --
440	+
441	+	#[test]
442	+	fn empty_input() {
443	+	assert_eq!(convert(""), "");
444	+	}
445	+
446	+	#[test]
447	+	fn plain_text() {
448	+	assert_eq!(convert("hello world"), "hello world");
449	+	}
450	+
451	+	#[test]
452	+	fn paragraph() {
453	+	assert_eq!(convert("<p>one</p><p>two</p>"), "one\n\ntwo");
454	+	}
455	+
456	+	#[test]
457	+	fn headings() {
458	+	assert_eq!(convert("<h1>Title</h1>"), "# Title");
459	+	assert_eq!(convert("<h3>Sub</h3>"), "### Sub");
460	+	}
461	+
462	+	#[test]
463	+	fn bold_and_italic() {
464	+	assert_eq!(
465	+	convert("<p><strong>bold</strong> and <em>italic</em></p>"),
466	+	"bold and italic"
467	+	);
468	+	}
469	+
470	+	#[test]
471	+	fn link() {
472	+	assert_eq!(
473	+	convert(r#"<a href="https://example.com">click</a>"#),
474	+	"[click](https://example.com)"
475	+	);
476	+	}
477	+
478	+	#[test]
479	+	fn link_text_matches_url() {
480	+	assert_eq!(
481	+	convert(r#"<a href="https://example.com">https://example.com</a>"#),
482	+	"https://example.com"
483	+	);
484	+	}
485	+
486	+	#[test]
487	+	fn link_empty_href() {
488	+	assert_eq!(convert(r#"<a href="">click</a>"#), "click");
489	+	}
490	+
491	+	#[test]
492	+	fn image() {
493	+	assert_eq!(
494	+	convert(r#"<img src="photo.jpg" alt="A photo">"#),
495	+	"![A photo](photo.jpg)"
496	+	);
497	+	}
498	+
499	+	#[test]
500	+	fn tracking_pixel_skipped() {

Lines truncated

A src/elements.rs +156

		@@ -0,0 +1,156 @@
1	+	use scraper::node::Element;
2	+
3	+	/// What kind of markdown wrapper an element produces.
4	+	pub enum ElementAction {
5	+	/// Skip this element and all its children entirely.
6	+	Skip,
7	+	/// Render children only, no wrapper (transparent element).
8	+	Transparent,
9	+	/// Block element with specific rendering.
10	+	Block(BlockKind),
11	+	/// Inline element with specific rendering.
12	+	Inline(InlineKind),
13	+	}
14	+
15	+	pub enum BlockKind {
16	+	Paragraph,
17	+	Heading(u8),
18	+	Blockquote,
19	+	UnorderedList,
20	+	OrderedList,
21	+	ListItem,
22	+	PreFormatted,
23	+	HorizontalRule,
24	+	Table,
25	+	Div,
26	+	}
27	+
28	+	pub enum InlineKind {
29	+	Bold,
30	+	Italic,
31	+	Strikethrough,
32	+	Code,
33	+	Link,
34	+	Image,
35	+	LineBreak,
36	+	Superscript,
37	+	Subscript,
38	+	}
39	+
40	+	/// Classify an HTML element into the action pter should take.
41	+	pub fn classify(el: &Element) -> ElementAction {
42	+	match el.name() {
43	+	// Skip entirely
44	+	"script" \| "style" \| "head" \| "meta" \| "link" \| "title" \| "noscript" => {
45	+	ElementAction::Skip
46	+	}
47	+
48	+	// Block elements
49	+	"p" => ElementAction::Block(BlockKind::Paragraph),
50	+	"h1" => ElementAction::Block(BlockKind::Heading(1)),
51	+	"h2" => ElementAction::Block(BlockKind::Heading(2)),
52	+	"h3" => ElementAction::Block(BlockKind::Heading(3)),
53	+	"h4" => ElementAction::Block(BlockKind::Heading(4)),
54	+	"h5" => ElementAction::Block(BlockKind::Heading(5)),
55	+	"h6" => ElementAction::Block(BlockKind::Heading(6)),
56	+	"blockquote" => ElementAction::Block(BlockKind::Blockquote),
57	+	"ul" \| "menu" => ElementAction::Block(BlockKind::UnorderedList),
58	+	"ol" => ElementAction::Block(BlockKind::OrderedList),
59	+	"li" => ElementAction::Block(BlockKind::ListItem),
60	+	"pre" => ElementAction::Block(BlockKind::PreFormatted),
61	+	"hr" => ElementAction::Block(BlockKind::HorizontalRule),
62	+	"table" => ElementAction::Block(BlockKind::Table),
63	+	// Table sub-elements are handled by the Table block handler, not individually
64	+	"thead" \| "tbody" \| "tfoot" \| "tr" \| "td" \| "th" \| "caption" \| "colgroup" \| "col" => {
65	+	ElementAction::Transparent
66	+	}
67	+	"div" \| "section" \| "article" \| "main" \| "header" \| "footer" \| "nav" \| "aside"
68	+	\| "figure" \| "figcaption" \| "details" \| "summary" => {
69	+	ElementAction::Block(BlockKind::Div)
70	+	}
71	+
72	+	// Inline elements
73	+	"strong" \| "b" => ElementAction::Inline(InlineKind::Bold),
74	+	"em" \| "i" => ElementAction::Inline(InlineKind::Italic),
75	+	"del" \| "s" \| "strike" => ElementAction::Inline(InlineKind::Strikethrough),
76	+	"code" \| "tt" => ElementAction::Inline(InlineKind::Code),
77	+	"a" => ElementAction::Inline(InlineKind::Link),
78	+	"img" => ElementAction::Inline(InlineKind::Image),
79	+	"br" => ElementAction::Inline(InlineKind::LineBreak),
80	+	"sup" => ElementAction::Inline(InlineKind::Superscript),
81	+	"sub" => ElementAction::Inline(InlineKind::Subscript),
82	+
83	+	// Everything else: transparent (render children)
84	+	_ => ElementAction::Transparent,
85	+	}
86	+	}
87	+
88	+	/// Check if an <img> element is a tracking pixel.
89	+	/// Returns true if it should be skipped.
90	+	pub fn is_tracking_pixel(el: &Element) -> bool {
91	+	let width = el.attr("width");
92	+	let height = el.attr("height");
93	+
94	+	// 1x1 or 0x0 images
95	+	if matches!(width, Some("1" \| "0")) \|\| matches!(height, Some("1" \| "0")) {
96	+	return true;
97	+	}
98	+
99	+	// No src attribute
100	+	let Some(src) = el.attr("src") else {
101	+	return true;
102	+	};
103	+
104	+	// Empty or data:image/gif (common transparent pixel)
105	+	if src.is_empty() {
106	+	return true;
107	+	}
108	+	if src.starts_with("data:image/gif;base64,R0lGOD") {
109	+	return true;
110	+	}
111	+
112	+	// Check inline style for tiny dimensions
113	+	if let Some(style) = el.attr("style") {
114	+	let style_lower = style.to_lowercase();
115	+	if style_lower.contains("width:1px")
116	+	\|\| style_lower.contains("width: 1px")
117	+	\|\| style_lower.contains("width:0")
118	+	\|\| style_lower.contains("height:1px")
119	+	\|\| style_lower.contains("height: 1px")
120	+	\|\| style_lower.contains("height:0")
121	+	\|\| style_lower.contains("display:none")
122	+	\|\| style_lower.contains("display: none")
123	+	{
124	+	return true;
125	+	}
126	+	}
127	+
128	+	false
129	+	}
130	+
131	+	/// Check if an element is hidden via inline style.
132	+	///
133	+	/// Catches display:none, visibility:hidden, and spacer tricks
134	+	/// like font-size:0 or line-height:0 (commonly used in email templates).
135	+	pub fn is_hidden(el: &Element) -> bool {
136	+	if let Some(style) = el.attr("style") {
137	+	let s = style.to_lowercase();
138	+	if s.contains("display:none")
139	+	\|\| s.contains("display: none")
140	+	\|\| s.contains("visibility:hidden")
141	+	\|\| s.contains("visibility: hidden")
142	+	\|\| s.contains("font-size:0")
143	+	\|\| s.contains("font-size: 0")
144	+	\|\| s.contains("line-height:0")
145	+	\|\| s.contains("line-height: 0")
146	+	\|\| (s.contains("height:0") && s.contains("overflow:hidden"))
147	+	\|\| (s.contains("height: 0") && s.contains("overflow: hidden"))
148	+	\|\| s.contains("max-height:0")
149	+	\|\| s.contains("max-height: 0")
150	+	{
151	+	return true;
152	+	}
153	+	}
154	+	false
155	+	}
156	+

A src/lib.rs +11

		@@ -0,0 +1,11 @@
1	+	//! pter — Plain Text Email Renderer
2	+	//!
3	+	//! Converts HTML email bodies into readable markdown.
4	+
5	+	mod convert;
6	+	mod elements;
7	+	mod replies;
8	+	mod tables;
9	+	mod whitespace;
10	+
11	+	pub use convert::convert;

A src/replies.rs +324

		@@ -0,0 +1,324 @@
1	+	use scraper::node::Node;
2	+	use scraper::ElementRef;
3	+
4	+	/// Check if an element marks the beginning of a quoted reply.
5	+	///
6	+	/// This is the central abstraction for reply detection. Rather than
7	+	/// building per-client logic throughout the converter, all client-specific
8	+	/// knowledge lives here behind a single predicate.
9	+	///
10	+	/// An element is a reply boundary if it's a container that wraps quoted
11	+	/// content from a previous message in the thread. The converter treats
12	+	/// these identically to `<blockquote>` — children get `>` prefixed.
13	+	pub fn is_reply_boundary(el: ElementRef) -> bool {
14	+	let element = el.value();
15	+	let name = element.name();
16	+
17	+	// <blockquote> is already handled by the element classifier.
18	+	// This function catches non-blockquote reply wrappers.
19	+
20	+	// Structural: elements with type="cite" (Apple Mail, some webmail)
21	+	if element.attr("type") == Some("cite") {
22	+	return true;
23	+	}
24	+
25	+	// Class/ID-based detection — thin per-client checks
26	+	if element.attr("class").is_some_and(is_reply_class) {
27	+	return true;
28	+	}
29	+
30	+	if element.attr("id").is_some_and(is_reply_id) {
31	+	return true;
32	+	}
33	+
34	+	// Heuristic: a <div> whose first meaningful text child matches
35	+	// an attribution pattern ("On ... wrote:") followed by a blockquote
36	+	if name == "div" && has_attribution_then_quote(el) {
37	+	return true;
38	+	}
39	+
40	+	false
41	+	}
42	+
43	+	/// Extract an attribution line from just before or at the start of a reply boundary.
44	+	///
45	+	/// Returns the attribution text (e.g. "On Mon, Jan 5, Alice wrote:") if found,
46	+	/// so the converter can render it above the quoted block.
47	+	pub fn find_attribution(el: ElementRef) -> Option<String> {
48	+	// Check the element's own leading text for attribution patterns
49	+	for child in el.children() {
50	+	match child.value() {
51	+	Node::Text(text) => {
52	+	let trimmed = text.text.trim();
53	+	if is_attribution_text(trimmed) {
54	+	return Some(trimmed.to_string());
55	+	}
56	+	}
57	+	Node::Element(_) => {
58	+	// Stop at the first child element — attribution is leading text
59	+	break;
60	+	}
61	+	_ => {}
62	+	}
63	+	}
64	+
65	+	// Check for a preceding sibling text node or element with attribution
66	+	if let Some(prev) = previous_sibling_text(el) {
67	+	let trimmed = prev.trim().to_string();
68	+	if is_attribution_text(&trimmed) {
69	+	return Some(trimmed);
70	+	}
71	+	}
72	+
73	+	None
74	+	}
75	+
76	+	/// Check if text matches common email attribution patterns.
77	+	///
78	+	/// These patterns are cross-client — every email client generates some
79	+	/// variant of "On [date], [person] wrote:" or "--- Forwarded message ---".
80	+	fn is_attribution_text(text: &str) -> bool {
81	+	let t = text.trim();
82	+
83	+	// "On ... wrote:" (Gmail, Apple Mail, Thunderbird, most clients)
84	+	if t.starts_with("On ") && t.ends_with("wrote:") {
85	+	return true;
86	+	}
87	+	// Localized variants: "Le ... a écrit :" (French), "Am ... schrieb" (German)
88	+	if (t.starts_with("Le ") \|\| t.starts_with("El "))
89	+	&& (t.ends_with("crit :") \|\| t.ends_with("crit:"))
90	+	{
91	+	return true;
92	+	}
93	+	if t.starts_with("Am ") && (t.ends_with("schrieb:") \|\| t.ends_with("schrieb :")) {
94	+	return true;
95	+	}
96	+
97	+	// Forwarded message separators
98	+	if t.contains("Forwarded message")
99	+	\|\| t.contains("Begin forwarded message")
100	+	\|\| t.contains("Original Message")
101	+	{
102	+	return true;
103	+	}
104	+
105	+	false
106	+	}
107	+
108	+	/// Thin per-client class checks. Each is one line — easy to add new clients.
109	+	fn is_reply_class(class: &str) -> bool {
110	+	// Split on whitespace to check individual class names
111	+	class.split_whitespace().any(\|c\| {
112	+	matches!(
113	+	c,
114	+	"gmail_quote"
115	+	\| "gmail_extra"
116	+	\| "yahoo_quoted"
117	+	\| "protonmail_quote"
118	+	\| "tutanota_quote"
119	+	\| "moz-cite-prefix" // Thunderbird
120	+	\| "zmail_extra" // Zoho
121	+	\| "WordSection1" // Outlook (sometimes wraps replies)
122	+	)
123	+	})
124	+	}
125	+
126	+	/// Thin per-client ID checks.
127	+	fn is_reply_id(id: &str) -> bool {
128	+	matches!(
129	+	id,
130	+	"divRplyFwdMsg" // Outlook
131	+	\| "reply-message" // Generic
132	+	\| "OLK_SRC_BODY_SECTION" // Outlook Mac
133	+	)
134	+	}
135	+
136	+	/// Check if a div contains attribution text followed by a blockquote.
137	+	///
138	+	/// This catches the common pattern where no class/id is present but
139	+	/// the structure is: `<div>On ... wrote:<br><blockquote>...</blockquote></div>`
140	+	fn has_attribution_then_quote(el: ElementRef) -> bool {
141	+	let mut found_attribution = false;
142	+
143	+	for child in el.children() {
144	+	match child.value() {
145	+	Node::Text(text) => {
146	+	if is_attribution_text(text.text.trim()) {
147	+	found_attribution = true;
148	+	}
149	+	}
150	+	Node::Element(e) => {
151	+	if found_attribution && e.name() == "blockquote" {
152	+	return true;
153	+	}
154	+	// Skip <br> tags between attribution and blockquote
155	+	if e.name() != "br" {
156	+	// If we hit a non-br element before finding attribution, stop
157	+	if !found_attribution {
158	+	return false;
159	+	}
160	+	}
161	+	}
162	+	_ => {}
163	+	}
164	+	}
165	+
166	+	false
167	+	}
168	+
169	+	/// Get text from the previous sibling, if it exists and is a text or inline element.
170	+	fn previous_sibling_text(el: ElementRef) -> Option<String> {
171	+	let prev = el.prev_sibling()?;
172	+
173	+	match prev.value() {
174	+	Node::Text(text) => Some(text.text.to_string()),
175	+	Node::Element(e) => {
176	+	// Check inline elements like <span>, <font> that might wrap attribution
177	+	if matches!(e.name(), "span" \| "font" \| "b" \| "i" \| "div" \| "p") {
178	+	let el_ref = ElementRef::wrap(prev)?;
179	+	let text: String = el_ref.text().collect();
180	+	if !text.trim().is_empty() {
181	+	return Some(text);
182	+	}
183	+	}
184	+	None
185	+	}
186	+	_ => None,
187	+	}
188	+	}
189	+
190	+	/// Check if a separator element marks the boundary between original
191	+	/// content and a forwarded/replied message.
192	+	///
193	+	/// This catches `<hr>` or styled divs that act as visual separators
194	+	/// before reply content (common in Outlook "From: ... Sent: ..." blocks).
195	+	pub fn is_outlook_separator(el: ElementRef) -> bool {
196	+	let element = el.value();
197	+
198	+	// Outlook uses a specific pattern: a div containing
199	+	// "From: ... Sent: ... To: ... Subject: ..." as a reply header
200	+	if element.name() == "div" \|\| element.name() == "p" {
201	+	let text: String = el.text().collect();
202	+	let t = text.trim();
203	+
204	+	// Must have at least From + Sent/Date or Subject
205	+	let has_from = t.contains("From:");
206	+	let has_sent = t.contains("Sent:") \|\| t.contains("Date:");
207	+	let has_subject = t.contains("Subject:");
208	+
209	+	if has_from && (has_sent \|\| has_subject) {
210	+	return true;
211	+	}
212	+	}
213	+
214	+	false
215	+	}
216	+
217	+	#[cfg(test)]
218	+	mod tests {
219	+	use super::*;
220	+	use scraper::{Html, Selector};
221	+
222	+	fn parse_and_select(html: &str, selector: &str) -> (Html, Selector) {
223	+	let doc = Html::parse_document(html);
224	+	let sel = Selector::parse(selector).unwrap();
225	+	(doc, sel)
226	+	}
227	+
228	+	// -- Attribution detection --
229	+
230	+	#[test]
231	+	fn attribution_on_wrote() {
232	+	assert!(is_attribution_text("On Mon, Jan 5, 2026 at 3:00 PM Alice <alice@example.com> wrote:"));
233	+	}
234	+
235	+	#[test]
236	+	fn attribution_forwarded() {
237	+	assert!(is_attribution_text("---------- Forwarded message ----------"));
238	+	}
239	+
240	+	#[test]
241	+	fn attribution_original_message() {
242	+	assert!(is_attribution_text("-----Original Message-----"));
243	+	}
244	+
245	+	#[test]
246	+	fn attribution_begin_forwarded() {
247	+	assert!(is_attribution_text("Begin forwarded message:"));
248	+	}
249	+
250	+	#[test]
251	+	fn not_attribution() {
252	+	assert!(!is_attribution_text("Hello, how are you?"));
253	+	assert!(!is_attribution_text("On the other hand, this is fine."));
254	+	}
255	+
256	+	// -- Reply class detection --
257	+
258	+	#[test]
259	+	fn gmail_quote_class() {
260	+	assert!(is_reply_class("gmail_quote"));
261	+	}
262	+
263	+	#[test]
264	+	fn multiple_classes_with_reply() {
265	+	assert!(is_reply_class("some-class gmail_quote another"));
266	+	}
267	+
268	+	#[test]
269	+	fn non_reply_class() {
270	+	assert!(!is_reply_class("regular-div content-wrapper"));
271	+	}
272	+
273	+	// -- Reply boundary detection --
274	+
275	+	#[test]
276	+	fn type_cite_is_boundary() {
277	+	let html = r#"<div type="cite"><p>quoted</p></div>"#;
278	+	let (doc, sel) = parse_and_select(html, r#"div[type="cite"]"#);
279	+	let el = doc.select(&sel).next().unwrap();
280	+	assert!(is_reply_boundary(el));
281	+	}
282	+
283	+	#[test]
284	+	fn gmail_quote_is_boundary() {
285	+	let html = r#"<div class="gmail_quote"><p>quoted</p></div>"#;
286	+	let (doc, sel) = parse_and_select(html, "div.gmail_quote");
287	+	let el = doc.select(&sel).next().unwrap();
288	+	assert!(is_reply_boundary(el));
289	+	}
290	+
291	+	#[test]
292	+	fn outlook_id_is_boundary() {
293	+	let html = r#"<div id="divRplyFwdMsg"><p>quoted</p></div>"#;
294	+	let (doc, sel) = parse_and_select(html, "#divRplyFwdMsg");
295	+	let el = doc.select(&sel).next().unwrap();
296	+	assert!(is_reply_boundary(el));
297	+	}
298	+
299	+	#[test]
300	+	fn plain_div_not_boundary() {
301	+	let html = r#"<div class="content"><p>not quoted</p></div>"#;
302	+	let (doc, sel) = parse_and_select(html, "div.content");
303	+	let el = doc.select(&sel).next().unwrap();
304	+	assert!(!is_reply_boundary(el));
305	+	}
306	+
307	+	// -- Outlook separator --
308	+
309	+	#[test]
310	+	fn outlook_from_sent_subject() {
311	+	let html = "<div>From: Alice\nSent: Monday\nTo: Bob\nSubject: Hello</div>";
312	+	let (doc, sel) = parse_and_select(html, "div");
313	+	let el = doc.select(&sel).next().unwrap();
314	+	assert!(is_outlook_separator(el));
315	+	}
316	+
317	+	#[test]
318	+	fn regular_div_not_separator() {
319	+	let html = "<div>Just a normal paragraph.</div>";
320	+	let (doc, sel) = parse_and_select(html, "div");
321	+	let el = doc.select(&sel).next().unwrap();
322	+	assert!(!is_outlook_separator(el));
323	+	}
324	+	}

A src/tables.rs +297

		@@ -0,0 +1,297 @@
1	+	use scraper::ElementRef;
2	+
3	+	/// Determine whether a `<table>` element is a data table or a layout table.
4	+	///
5	+	/// Email HTML overwhelmingly uses tables for layout. A table is considered
6	+	/// a data table if it has structural indicators of tabular data:
7	+	/// - Contains `<th>` elements
8	+	/// - Has a `<caption>` child
9	+	/// - Has `role="grid"` or `role="table"`
10	+	/// - Has multiple rows where multiple cells contain substantive text
11	+	///
12	+	/// Everything else is treated as a layout table and unwrapped.
13	+	pub fn is_data_table(table: ElementRef) -> bool {
14	+	let el = table.value();
15	+
16	+	// role attribute
17	+	if let Some(role) = el.attr("role") {
18	+	if role == "grid" \|\| role == "table" {
19	+	return true;
20	+	}
21	+	// role="presentation" is an explicit layout signal
22	+	if role == "presentation" \|\| role == "none" {
23	+	return false;
24	+	}
25	+	}
26	+
27	+	let mut has_th = false;
28	+	let mut has_caption = false;
29	+	let mut multi_cell_rows = 0u32;
30	+
31	+	for descendant in table.descendants() {
32	+	if let Some(el_ref) = ElementRef::wrap(descendant) {
33	+	match el_ref.value().name() {
34	+	"th" => has_th = true,
35	+	"caption" => has_caption = true,
36	+	"tr" => {
37	+	let cell_count = el_ref
38	+	.children()
39	+	.filter_map(ElementRef::wrap)
40	+	.filter(\|c\| {
41	+	let name = c.value().name();
42	+	(name == "td" \|\| name == "th") && has_substantive_text(*c)
43	+	})
44	+	.count();
45	+	if cell_count > 1 {
46	+	multi_cell_rows += 1;
47	+	}
48	+	}
49	+	_ => {}
50	+	}
51	+	}
52	+	}
53	+
54	+	if has_th \|\| has_caption {
55	+	return true;
56	+	}
57	+
58	+	// Multiple rows with multiple substantive cells = data table
59	+	multi_cell_rows >= 2
60	+	}
61	+
62	+	/// Check if an element contains meaningful text (not just whitespace/nbsp).
63	+	fn has_substantive_text(el: ElementRef) -> bool {
64	+	let text = el.text().collect::<String>();
65	+	let trimmed = text.trim().replace('\u{a0}', ""); // strip
66	+	trimmed.len() > 1 // more than a single character
67	+	}
68	+
69	+	/// Extract rows and cells from a data table for markdown rendering.
70	+	///
71	+	/// Returns (headers, rows) where each is a Vec of cell text strings.
72	+	/// If no `<thead>`/`<th>` row exists, the first row is used as headers.
73	+	pub fn extract_table_data(table: ElementRef) -> (Vec<String>, Vec<Vec<String>>) {
74	+	let mut headers: Vec<String> = Vec::new();
75	+	let mut rows: Vec<Vec<String>> = Vec::new();
76	+
77	+	// Look for thead/th first
78	+	for descendant in table.children().filter_map(ElementRef::wrap) {
79	+	let name = descendant.value().name();
80	+	if name == "thead" {
81	+	for tr in descendant.children().filter_map(ElementRef::wrap) {
82	+	if tr.value().name() == "tr" {
83	+	headers = extract_cells(tr);
84	+	break; // first row of thead
85	+	}
86	+	}
87	+	} else if name == "tbody" \|\| name == "tr" {
88	+	let trs: Box<dyn Iterator<Item = ElementRef>> = if name == "tbody" {
89	+	Box::new(
90	+	descendant
91	+	.children()
92	+	.filter_map(ElementRef::wrap)
93	+	.filter(\|e\| e.value().name() == "tr"),
94	+	)
95	+	} else {
96	+	Box::new(std::iter::once(descendant))
97	+	};
98	+
99	+	for tr in trs {
100	+	let cells = extract_cells(tr);
101	+	if !cells.is_empty() {
102	+	// If we haven't found headers yet and this row has <th> cells,
103	+	// treat it as the header row
104	+	if headers.is_empty() && has_th_cells(tr) {
105	+	headers = cells;
106	+	} else {
107	+	rows.push(cells);
108	+	}
109	+	}
110	+	}
111	+	}
112	+	}
113	+
114	+	// If still no headers, promote first data row
115	+	if headers.is_empty() && !rows.is_empty() {
116	+	headers = rows.remove(0);
117	+	}
118	+
119	+	(headers, rows)
120	+	}
121	+
122	+	fn extract_cells(tr: ElementRef) -> Vec<String> {
123	+	tr.children()
124	+	.filter_map(ElementRef::wrap)
125	+	.filter(\|e\| {
126	+	let n = e.value().name();
127	+	n == "td" \|\| n == "th"
128	+	})
129	+	.map(\|cell\| {
130	+	let text = cell.text().collect::<String>();
131	+	text.split_whitespace().collect::<Vec<_>>().join(" ")
132	+	})
133	+	.collect()
134	+	}
135	+
136	+	fn has_th_cells(tr: ElementRef) -> bool {
137	+	tr.children()
138	+	.filter_map(ElementRef::wrap)
139	+	.any(\|e\| e.value().name() == "th")
140	+	}
141	+
142	+	/// Render a data table as a GFM markdown table.
143	+	pub fn render_markdown_table(headers: &[String], rows: &[Vec<String>]) -> String {
144	+	if headers.is_empty() {
145	+	return String::new();
146	+	}
147	+
148	+	let col_count = headers.len();
149	+	let mut out = String::new();
150	+
151	+	// Header row
152	+	out.push('\|');
153	+	for h in headers {
154	+	out.push(' ');
155	+	out.push_str(h);
156	+	out.push_str(" \|");
157	+	}
158	+	out.push('\n');
159	+
160	+	// Separator row
161	+	out.push('\|');
162	+	for _ in 0..col_count {
163	+	out.push_str(" --- \|");
164	+	}
165	+	out.push('\n');
166	+
167	+	// Data rows
168	+	for row in rows {
169	+	out.push('\|');
170	+	for i in 0..col_count {
171	+	out.push(' ');
172	+	if let Some(cell) = row.get(i) {
173	+	out.push_str(cell);
174	+	}
175	+	out.push_str(" \|");
176	+	}
177	+	out.push('\n');
178	+	}
179	+
180	+	// Remove trailing newline (caller handles spacing)
181	+	out.trim_end().to_string()
182	+	}
183	+
184	+	#[cfg(test)]
185	+	mod tests {
186	+	use super::*;
187	+	use scraper::{Html, Selector};
188	+
189	+	fn parse_table(html: &str) -> Html {
190	+	Html::parse_document(html)
191	+	}
192	+
193	+	fn select_table(doc: &Html) -> ElementRef<'_> {
194	+	let sel = Selector::parse("table").unwrap();
195	+	doc.select(&sel).next().unwrap()
196	+	}
197	+
198	+	#[test]
199	+	fn single_cell_is_layout() {
200	+	let doc = parse_table("<table><tr><td>content</td></tr></table>");
201	+	assert!(!is_data_table(select_table(&doc)));
202	+	}
203	+
204	+	#[test]
205	+	fn table_with_th_is_data() {
206	+	let doc = parse_table(
207	+	"<table><tr><th>Name</th><th>Age</th></tr><tr><td>Alice</td><td>30</td></tr></table>",
208	+	);
209	+	assert!(is_data_table(select_table(&doc)));
210	+	}
211	+
212	+	#[test]
213	+	fn table_with_caption_is_data() {
214	+	let doc = parse_table(
215	+	"<table><caption>Users</caption><tr><td>Alice</td><td>30</td></tr></table>",
216	+	);
217	+	assert!(is_data_table(select_table(&doc)));
218	+	}
219	+
220	+	#[test]
221	+	fn role_presentation_is_layout() {
222	+	let doc = parse_table(
223	+	r#"<table role="presentation"><tr><td>layout</td><td>stuff</td></tr></table>"#,
224	+	);
225	+	assert!(!is_data_table(select_table(&doc)));
226	+	}
227	+
228	+	#[test]
229	+	fn role_grid_is_data() {
230	+	let doc =
231	+	parse_table(r#"<table role="grid"><tr><td>Alice</td><td>30</td></tr></table>"#);
232	+	assert!(is_data_table(select_table(&doc)));
233	+	}
234	+
235	+	#[test]
236	+	fn multi_row_multi_cell_is_data() {
237	+	let doc = parse_table(
238	+	"<table>\
239	+	<tr><td>Alice</td><td>Engineer</td></tr>\
240	+	<tr><td>Bob</td><td>Designer</td></tr>\
241	+	</table>",
242	+	);
243	+	assert!(is_data_table(select_table(&doc)));
244	+	}
245	+
246	+	#[test]
247	+	fn spacer_cells_not_substantive() {
248	+	let doc = parse_table(
249	+	"<table><tr><td>content</td><td> </td></tr>\
250	+	<tr><td>more</td><td> </td></tr></table>",
251	+	);
252	+	// Only one substantive cell per row
253	+	assert!(!is_data_table(select_table(&doc)));
254	+	}
255	+
256	+	#[test]
257	+	fn render_simple_table() {
258	+	let headers = vec!["Name".into(), "Age".into()];
259	+	let rows = vec![
260	+	vec!["Alice".into(), "30".into()],
261	+	vec!["Bob".into(), "25".into()],
262	+	];
263	+	let md = render_markdown_table(&headers, &rows);
264	+	assert_eq!(
265	+	md,
266	+	"\| Name \| Age \|\n\| --- \| --- \|\n\| Alice \| 30 \|\n\| Bob \| 25 \|"
267	+	);
268	+	}
269	+
270	+	#[test]
271	+	fn render_empty_headers() {
272	+	let md = render_markdown_table(&[], &[]);
273	+	assert_eq!(md, "");
274	+	}
275	+
276	+	#[test]
277	+	fn extract_with_thead() {
278	+	let doc = parse_table(
279	+	"<table><thead><tr><th>A</th><th>B</th></tr></thead>\
280	+	<tbody><tr><td>1</td><td>2</td></tr></tbody></table>",
281	+	);
282	+	let (h, r) = extract_table_data(select_table(&doc));
283	+	assert_eq!(h, vec!["A", "B"]);
284	+	assert_eq!(r, vec![vec!["1".to_string(), "2".to_string()]]);
285	+	}
286	+
287	+	#[test]
288	+	fn extract_promotes_first_row() {
289	+	let doc = parse_table(
290	+	"<table><tr><td>Name</td><td>Val</td></tr>\
291	+	<tr><td>X</td><td>Y</td></tr></table>",
292	+	);
293	+	let (h, r) = extract_table_data(select_table(&doc));
294	+	assert_eq!(h, vec!["Name", "Val"]);
295	+	assert_eq!(r, vec![vec!["X".to_string(), "Y".to_string()]]);
296	+	}
297	+	}

A src/whitespace.rs +56

		@@ -0,0 +1,56 @@
1	+	/// Normalize whitespace in the final markdown output.
2	+	///
3	+	/// - Collapse runs of 3+ newlines into 2 (one blank line)
4	+	/// - Trim leading/trailing whitespace
5	+	/// - Remove trailing whitespace from each line
6	+	pub fn normalize(input: &str) -> String {
7	+	let mut result = String::with_capacity(input.len());
8	+	let mut consecutive_newlines = 0u32;
9	+
10	+	for ch in input.chars() {
11	+	if ch == '\n' {
12	+	consecutive_newlines += 1;
13	+	if consecutive_newlines <= 2 {
14	+	result.push('\n');
15	+	}
16	+	} else {
17	+	consecutive_newlines = 0;
18	+	result.push(ch);
19	+	}
20	+	}
21	+
22	+	// Trim trailing whitespace from each line
23	+	let lines: Vec<&str> = result.lines().map(\|l\| l.trim_end()).collect();
24	+	let joined = lines.join("\n");
25	+	joined.trim().to_string()
26	+	}
27	+
28	+	#[cfg(test)]
29	+	mod tests {
30	+	use super::*;
31	+
32	+	#[test]
33	+	fn collapse_excessive_newlines() {
34	+	assert_eq!(normalize("a\n\n\n\nb"), "a\n\nb");
35	+	}
36	+
37	+	#[test]
38	+	fn preserve_single_blank_line() {
39	+	assert_eq!(normalize("a\n\nb"), "a\n\nb");
40	+	}
41	+
42	+	#[test]
43	+	fn trim_trailing_whitespace() {
44	+	assert_eq!(normalize("hello \nworld "), "hello\nworld");
45	+	}
46	+
47	+	#[test]
48	+	fn trim_outer_whitespace() {
49	+	assert_eq!(normalize("\n\nhello\n\n"), "hello");
50	+	}
51	+
52	+	#[test]
53	+	fn empty_input() {
54	+	assert_eq!(normalize(""), "");
55	+	}
56	+	}

A tests/edge_cases.rs +202

		@@ -0,0 +1,202 @@
1	+	use pter::convert;
2	+
3	+	#[test]
4	+	fn empty_string() {
5	+	assert_eq!(convert(""), "");
6	+	}
7	+
8	+	#[test]
9	+	fn whitespace_only() {
10	+	assert_eq!(convert(" \n\t "), "");
11	+	}
12	+
13	+	#[test]
14	+	fn just_tags_no_content() {
15	+	assert_eq!(convert("<div><p><span></span></p></div>"), "");
16	+	}
17	+
18	+	#[test]
19	+	fn deeply_nested_divs() {
20	+	let mut html = String::new();
21	+	for _ in 0..100 {
22	+	html.push_str("<div>");
23	+	}
24	+	html.push_str("deep content");
25	+	for _ in 0..100 {
26	+	html.push_str("</div>");
27	+	}
28	+	let md = convert(&html);
29	+	assert!(md.contains("deep content"));
30	+	}
31	+
32	+	#[test]
33	+	fn deeply_nested_blockquotes() {
34	+	let mut html = String::new();
35	+	for _ in 0..20 {
36	+	html.push_str("<blockquote>");
37	+	}
38	+	html.push_str("very deep");
39	+	for _ in 0..20 {
40	+	html.push_str("</blockquote>");
41	+	}
42	+	let md = convert(&html);
43	+	assert!(md.contains("very deep"));
44	+	// Should have many > prefixes
45	+	assert!(md.contains("> > > > >"));
46	+	}
47	+
48	+	#[test]
49	+	fn deeply_nested_lists() {
50	+	let mut html = String::new();
51	+	for _ in 0..10 {
52	+	html.push_str("<ul><li>");
53	+	}
54	+	html.push_str("deep item");
55	+	for _ in 0..10 {
56	+	html.push_str("</li></ul>");
57	+	}
58	+	let md = convert(&html);
59	+	assert!(md.contains("deep item"));
60	+	}
61	+
62	+	#[test]
63	+	fn malformed_unclosed_tags() {
64	+	// html5ever auto-corrects these
65	+	let md = convert("<p>unclosed paragraph<p>another one");
66	+	assert!(md.contains("unclosed paragraph"));
67	+	assert!(md.contains("another one"));
68	+	}
69	+
70	+	#[test]
71	+	fn malformed_mismatched_tags() {
72	+	let md = convert("<b><i>crossed</b></i>");
73	+	assert!(md.contains("crossed"));
74	+	}
75	+
76	+	#[test]
77	+	fn only_script_content() {
78	+	assert_eq!(convert("<script>alert('xss')</script>"), "");
79	+	}
80	+
81	+	#[test]
82	+	fn only_style_content() {
83	+	assert_eq!(convert("<style>.x { color: red; }</style>"), "");
84	+	}
85	+
86	+	#[test]
87	+	fn only_tracking_pixels() {
88	+	let html = r#"
89	+	<img src="a.gif" width="1" height="1">
90	+	<img src="b.gif" width="1" height="1">
91	+	"#;
92	+	assert_eq!(convert(html), "");
93	+	}
94	+
95	+	#[test]
96	+	fn unicode_content() {
97	+	let md = convert("<p>日本語テスト 🎉 émojis café</p>");
98	+	assert!(md.contains("日本語テスト"));
99	+	assert!(md.contains("🎉"));
100	+	assert!(md.contains("café"));
101	+	}
102	+
103	+	#[test]
104	+	fn html_entities_numeric() {
105	+	let md = convert("<p>© — ’</p>");
106	+	assert!(md.contains("©"));
107	+	assert!(md.contains("—"));
108	+	}
109	+
110	+	#[test]
111	+	fn large_input_doesnt_blow_up() {
112	+	let para = "<p>Hello world. This is a test paragraph with some content.</p>";
113	+	let html: String = para.repeat(1000);
114	+	let md = convert(&html);
115	+	assert!(md.contains("Hello world"));
116	+	// Should be proportional, not quadratic
117	+	assert!(md.len() < html.len());
118	+	}
119	+
120	+	#[test]
121	+	fn link_with_nested_formatting() {
122	+	let html = r#"<a href="https://example.com"><strong>bold link</strong></a>"#;
123	+	let md = convert(html);
124	+	assert!(md.contains("[bold link](https://example.com)"));
125	+	}
126	+
127	+	#[test]
128	+	fn image_with_no_alt() {
129	+	let md = convert(r#"<img src="photo.jpg">"#);
130	+	assert!(md.contains("![](photo.jpg)"));
131	+	}
132	+
133	+	#[test]
134	+	fn consecutive_inline_elements() {
135	+	let md = convert("<b>bold</b><i>italic</i><code>code</code>");
136	+	assert_eq!(md, "bolditalic`code`");
137	+	}
138	+
139	+	#[test]
140	+	fn table_with_empty_cells() {
141	+	let html = "<table><tr><th>A</th><th>B</th></tr>\
142	+	<tr><td></td><td>val</td></tr></table>";
143	+	let md = convert(html);
144	+	assert!(md.contains("\| A \| B \|"));
145	+	assert!(md.contains("\| \| val \|"));
146	+	}
147	+
148	+	#[test]
149	+	fn pre_with_html_inside() {
150	+	let html = "<pre><div>not a tag</div></pre>";
151	+	let md = convert(html);
152	+	assert!(md.contains("```"));
153	+	assert!(md.contains("<div>not a tag</div>"));
154	+	}
155	+
156	+	#[test]
157	+	fn multiple_spaces_in_source() {
158	+	let md = convert("<p>word1 word2 word3</p>");
159	+	assert_eq!(md, "word1 word2 word3");
160	+	}
161	+
162	+	#[test]
163	+	fn newlines_in_source_collapsed() {
164	+	let md = convert("<p>line1\n\n\nline2</p>");
165	+	assert_eq!(md, "line1 line2");
166	+	}
167	+
168	+	#[test]
169	+	fn full_html_document() {
170	+	let html = r#"
171	+	<!DOCTYPE html>
172	+	<html lang="en">
173	+	<head>
174	+	<meta charset="UTF-8">
175	+	<title>Test Email</title>
176	+	<style>body { font-family: sans-serif; }</style>
177	+	</head>
178	+	<body>
179	+	<p>Hello!</p>
180	+	</body>
181	+	</html>
182	+	"#;
183	+	let md = convert(html);
184	+	assert_eq!(md, "Hello!");
185	+	}
186	+
187	+	#[test]
188	+	fn data_uri_image_not_tracking_pixel() {
189	+	// A data URI image that's not 1x1 should render
190	+	let html = r#"<img src="data:image/png;base64,iVBOR..." alt="inline" width="100">"#;
191	+	let md = convert(html);
192	+	assert!(md.contains("![inline]"));
193	+	}
194	+
195	+	#[test]
196	+	fn blockquote_with_paragraphs() {
197	+	let html = "<blockquote><p>First para</p><p>Second para</p></blockquote>";
198	+	let md = convert(html);
199	+	assert!(md.contains("> First para"));
200	+	assert!(md.contains("> "));
201	+	assert!(md.contains("> Second para"));
202	+	}

A tests/integration.rs +436

		@@ -0,0 +1,436 @@
1	+	use pter::convert;
2	+
3	+	#[test]
4	+	fn simple_email() {
5	+	let html = r#"
6	+	<html>
7	+	<head><title>Email</title></head>
8	+	<body>
9	+	<h1>Meeting Tomorrow</h1>
10	+	<p>Hi Max,</p>
11	+	<p>Just confirming our meeting tomorrow at <strong>2pm</strong>.</p>
12	+	<p>Best,<br>Alice</p>
13	+	</body>
14	+	</html>
15	+	"#;
16	+
17	+	let md = convert(html);
18	+	assert!(md.contains("# Meeting Tomorrow"));
19	+	assert!(md.contains("Hi Max,"));
20	+	assert!(md.contains("2pm"));
21	+	assert!(md.contains("Best,\nAlice"));
22	+	}
23	+
24	+	#[test]
25	+	fn email_with_links() {
26	+	let html = r#"
27	+	<body>
28	+	<p>Please review the <a href="https://example.com/doc">document</a>.</p>
29	+	<p>Direct link: <a href="https://example.com">https://example.com</a></p>
30	+	</body>
31	+	"#;
32	+
33	+	let md = convert(html);
34	+	assert!(md.contains("[document](https://example.com/doc)"));
35	+	// Link text matches URL — no markdown link syntax
36	+	assert!(md.contains("Direct link: https://example.com"));
37	+	}
38	+
39	+	#[test]
40	+	fn email_with_tracking_pixels() {
41	+	let html = r#"
42	+	<body>
43	+	<p>Content here</p>
44	+	<img src="https://tracker.example.com/open.gif" width="1" height="1" alt="">
45	+	<img src="data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7" alt="">
46	+	<img src="real-image.jpg" alt="A real photo" width="600">
47	+	</body>
48	+	"#;
49	+
50	+	let md = convert(html);
51	+	assert!(md.contains("Content here"));
52	+	assert!(!md.contains("tracker"));
53	+	assert!(!md.contains("data:image"));
54	+	assert!(md.contains("![A real photo](real-image.jpg)"));
55	+	}
56	+
57	+	#[test]
58	+	fn email_with_quoted_reply() {
59	+	let html = r#"
60	+	<body>
61	+	<p>Thanks, that works for me.</p>
62	+	<blockquote>
63	+	<p>Can we meet at 3pm instead?</p>
64	+	</blockquote>
65	+	</body>
66	+	"#;
67	+
68	+	let md = convert(html);
69	+	assert!(md.contains("Thanks, that works for me."));
70	+	assert!(md.contains("> Can we meet at 3pm instead?"));
71	+	}
72	+
73	+	#[test]
74	+	fn email_with_signature_line() {
75	+	let html = r#"
76	+	<body>
77	+	<p>See you then.</p>
78	+	<hr>
79	+	<p>Alice Smith</p>
80	+	<p>Engineering Lead</p>
81	+	</body>
82	+	"#;
83	+
84	+	let md = convert(html);
85	+	assert!(md.contains("See you then."));
86	+	assert!(md.contains("---"));
87	+	assert!(md.contains("Alice Smith"));
88	+	}
89	+
90	+	#[test]
91	+	fn deeply_nested_blockquotes() {
92	+	let html = r#"
93	+	<body>
94	+	<p>Got it.</p>
95	+	<blockquote>
96	+	<p>Sounds good.</p>
97	+	<blockquote>
98	+	<p>Can we reschedule?</p>
99	+	<blockquote>
100	+	<p>Original message here.</p>
101	+	</blockquote>
102	+	</blockquote>
103	+	</blockquote>
104	+	</body>
105	+	"#;
106	+
107	+	let md = convert(html);
108	+	assert!(md.contains("Got it."));
109	+	assert!(md.contains("> Sounds good."));
110	+	assert!(md.contains("> > Can we reschedule?"));
111	+	assert!(md.contains("> > > Original message here."));
112	+	}
113	+
114	+	#[test]
115	+	fn complex_list_structure() {
116	+	let html = r#"
117	+	<body>
118	+	<p>Action items:</p>
119	+	<ol>
120	+	<li>Review the PR
121	+	<ul>
122	+	<li>Check tests</li>
123	+	<li>Check docs</li>
124	+	</ul>
125	+	</li>
126	+	<li>Deploy to staging</li>
127	+	</ol>
128	+	</body>
129	+	"#;
130	+
131	+	let md = convert(html);
132	+	assert!(md.contains("Action items:"));
133	+	assert!(md.contains("1. Review the PR"));
134	+	assert!(md.contains(" - Check tests"));
135	+	assert!(md.contains("2. Deploy to staging"));
136	+	}
137	+
138	+	#[test]
139	+	fn pre_block_preserves_formatting() {
140	+	let html = r#"
141	+	<body>
142	+	<p>Here's the code:</p>
143	+	<pre><code>fn main() {
144	+	println!("hello");
145	+	}</code></pre>
146	+	</body>
147	+	"#;
148	+
149	+	let md = convert(html);
150	+	assert!(md.contains("Here's the code:"));
151	+	assert!(md.contains("```\nfn main()"));
152	+	assert!(md.contains(" println!"));
153	+	}
154	+
155	+	#[test]
156	+	fn hidden_content_stripped() {
157	+	let html = r#"
158	+	<body>
159	+	<p>Visible content</p>
160	+	<div style="display: none;">
161	+	<p>This should not appear</p>
162	+	</div>
163	+	<span style="visibility: hidden;">Also hidden</span>
164	+	<p>More visible</p>
165	+	</body>
166	+	"#;
167	+
168	+	let md = convert(html);
169	+	assert!(md.contains("Visible content"));
170	+	assert!(!md.contains("should not appear"));
171	+	assert!(!md.contains("Also hidden"));
172	+	assert!(md.contains("More visible"));
173	+	}
174	+
175	+	#[test]
176	+	fn script_and_style_fully_removed() {
177	+	let html = r#"
178	+	<html>
179	+	<head>
180	+	<style>body { color: red; }</style>
181	+	<script>alert('xss');</script>
182	+	</head>
183	+	<body>
184	+	<p>Safe content</p>
185	+	<script>document.write('injected')</script>
186	+	</body>
187	+	</html>
188	+	"#;
189	+
190	+	let md = convert(html);
191	+	assert_eq!(md, "Safe content");
192	+	}
193	+
194	+	#[test]
195	+	fn newsletter_table_layout() {
196	+	// Typical email newsletter wrapped in layout tables
197	+	let html = r#"
198	+	<html>
199	+	<body>
200	+	<table width="100%" cellpadding="0" cellspacing="0" role="presentation">
201	+	<tr>
202	+	<td align="center">
203	+	<table width="600" cellpadding="0" cellspacing="0">
204	+	<tr>
205	+	<td>
206	+	<h2>Weekly Digest</h2>
207	+	<p>Here are your updates for this week.</p>
208	+	<ul>
209	+	<li>New release v2.0</li>
210	+	<li>Bug fixes</li>
211	+	</ul>
212	+	<p>Thanks for reading!</p>
213	+	</td>
214	+	</tr>
215	+	</table>
216	+	</td>
217	+	</tr>
218	+	</table>
219	+	<img src="https://track.example.com/open.gif" width="1" height="1">
220	+	</body>
221	+	</html>
222	+	"#;
223	+
224	+	let md = convert(html);
225	+	assert!(md.contains("## Weekly Digest"));
226	+	assert!(md.contains("Here are your updates for this week."));
227	+	assert!(md.contains("- New release v2.0"));
228	+	assert!(md.contains("- Bug fixes"));
229	+	assert!(md.contains("Thanks for reading!"));
230	+	assert!(!md.contains("track.example.com"));
231	+	// No table markup in output
232	+	assert!(!md.contains("\| "));
233	+	}
234	+
235	+	#[test]
236	+	fn data_table_preserved() {
237	+	let html = r#"
238	+	<body>
239	+	<p>Order summary:</p>
240	+	<table>
241	+	<thead><tr><th>Item</th><th>Qty</th><th>Price</th></tr></thead>
242	+	<tbody>
243	+	<tr><td>Widget</td><td>3</td><td>$15.00</td></tr>
244	+	<tr><td>Gadget</td><td>1</td><td>$29.99</td></tr>
245	+	</tbody>
246	+	</table>
247	+	</body>
248	+	"#;
249	+
250	+	let md = convert(html);
251	+	assert!(md.contains("Order summary:"));
252	+	assert!(md.contains("\| Item \| Qty \| Price \|"));
253	+	assert!(md.contains("\| --- \| --- \| --- \|"));
254	+	assert!(md.contains("\| Widget \| 3 \| $15.00 \|"));
255	+	assert!(md.contains("\| Gadget \| 1 \| $29.99 \|"));
256	+	}
257	+
258	+	#[test]
259	+	fn spacer_and_tracking_stripped() {
260	+	let html = r#"
261	+	<body>
262	+	<p>Real content</p>
263	+	<div style="font-size: 0; line-height: 0;"> </div>
264	+	<img src="pixel.gif" width="1" height="1" style="display:none">
265	+	<div style="height:0;overflow:hidden">invisible</div>
266	+	<p>More content</p>
267	+	</body>
268	+	"#;
269	+
270	+	let md = convert(html);
271	+	assert!(md.contains("Real content"));
272	+	assert!(md.contains("More content"));
273	+	assert!(!md.contains("invisible"));
274	+	assert!(!md.contains("pixel.gif"));
275	+	}
276	+
277	+	// -- Reply chain tests --
278	+
279	+	#[test]
280	+	fn gmail_reply_chain() {
281	+	let html = r#"
282	+	<body>
283	+	<div dir="ltr">
284	+	<p>Thanks, that works for me.</p>
285	+	</div>
286	+	<div class="gmail_quote">
287	+	<div class="gmail_attr">On Mon, Jan 5, 2026 at 3:00 PM Alice <alice@example.com> wrote:</div>
288	+	<blockquote class="gmail_quote">
289	+	<div dir="ltr">
290	+	<p>Can we meet at 3pm instead of 2pm?</p>
291	+	</div>
292	+	</blockquote>
293	+	</div>
294	+	</body>
295	+	"#;
296	+
297	+	let md = convert(html);
298	+	assert!(md.contains("Thanks, that works for me."));
299	+	// The gmail_quote div should be rendered as a quote block
300	+	assert!(md.contains("> "));
301	+	assert!(md.contains("3pm instead of 2pm"));
302	+	}
303	+
304	+	#[test]
305	+	fn apple_mail_reply() {
306	+	let html = r#"
307	+	<body>
308	+	<div>Sounds good, see you then.</div>
309	+	<div>
310	+	<br>
311	+	<blockquote type="cite">
312	+	<div>Hey, are we still on for lunch?</div>
313	+	</blockquote>
314	+	</div>
315	+	</body>
316	+	"#;
317	+
318	+	let md = convert(html);
319	+	assert!(md.contains("Sounds good, see you then."));
320	+	assert!(md.contains("> "));
321	+	assert!(md.contains("still on for lunch"));
322	+	}
323	+
324	+	#[test]
325	+	fn outlook_reply_with_separator() {
326	+	let html = r#"
327	+	<body>
328	+	<div>
329	+	<p>I'll handle it.</p>
330	+	</div>
331	+	<hr>
332	+	<div>
333	+	<p>From: Alice Smith<br>
334	+	Sent: Monday, January 5, 2026<br>
335	+	To: Bob Jones<br>
336	+	Subject: Action needed</p>
337	+	</div>
338	+	<div>
339	+	<p>Can you take a look at the report?</p>
340	+	</div>
341	+	</body>
342	+	"#;
343	+
344	+	let md = convert(html);
345	+	assert!(md.contains("I'll handle it."));
346	+	assert!(md.contains("---")); // hr separator
347	+	assert!(md.contains("From: Alice Smith"));
348	+	assert!(md.contains("take a look at the report"));
349	+	}
350	+
351	+	#[test]
352	+	fn nested_gmail_reply_chain() {
353	+	let html = r#"
354	+	<body>
355	+	<div dir="ltr"><p>Got it, thanks!</p></div>
356	+	<div class="gmail_quote">
357	+	On Tue, Jan 6, Bob wrote:
358	+	<blockquote class="gmail_quote">
359	+	<div dir="ltr"><p>Here's the update.</p></div>
360	+	<div class="gmail_quote">
361	+	On Mon, Jan 5, Alice wrote:
362	+	<blockquote class="gmail_quote">
363	+	<div dir="ltr"><p>What's the status?</p></div>
364	+	</blockquote>
365	+	</div>
366	+	</blockquote>
367	+	</div>
368	+	</body>
369	+	"#;
370	+
371	+	let md = convert(html);
372	+	assert!(md.contains("Got it, thanks!"));
373	+	// Should have nested quoting
374	+	assert!(md.contains("> "));
375	+	assert!(md.contains("Here's the update."));
376	+	assert!(md.contains("What's the status?"));
377	+	}
378	+
379	+	#[test]
380	+	fn forwarded_message() {
381	+	let html = r#"
382	+	<body>
383	+	<div><p>FYI, see below.</p></div>
384	+	<div class="gmail_quote">
385	+	---------- Forwarded message ----------
386	+	<blockquote>
387	+	<p>From: Alice</p>
388	+	<p>The deadline has been moved to Friday.</p>
389	+	</blockquote>
390	+	</div>
391	+	</body>
392	+	"#;
393	+
394	+	let md = convert(html);
395	+	assert!(md.contains("FYI, see below."));
396	+	assert!(md.contains("Forwarded message"));
397	+	assert!(md.contains("deadline has been moved"));
398	+	}
399	+
400	+	#[test]
401	+	fn protonmail_reply() {
402	+	let html = r#"
403	+	<body>
404	+	<div>Will do, thanks.</div>
405	+	<blockquote class="protonmail_quote" type="cite">
406	+	<div>Please send me the files by EOD.</div>
407	+	</blockquote>
408	+	</body>
409	+	"#;
410	+
411	+	let md = convert(html);
412	+	assert!(md.contains("Will do, thanks."));
413	+	assert!(md.contains("> "));
414	+	assert!(md.contains("send me the files"));
415	+	}
416	+
417	+	#[test]
418	+	fn attribution_preserved_above_quote() {
419	+	let html = r#"
420	+	<body>
421	+	<p>Agreed.</p>
422	+	<div class="gmail_quote">
423	+	On Wed, Jan 7, 2026 at 10:00 AM Carol wrote:
424	+	<blockquote>
425	+	<p>Let's go with option B.</p>
426	+	</blockquote>
427	+	</div>
428	+	</body>
429	+	"#;
430	+
431	+	let md = convert(html);
432	+	assert!(md.contains("Agreed."));
433	+	// Attribution should appear
434	+	assert!(md.contains("Carol wrote:"));
435	+	assert!(md.contains("option B"));
436	+	}

A tests/proptest.rs +94

		@@ -0,0 +1,94 @@
1	+	use proptest::prelude::*;
2	+
3	+	// Strategy: generate arbitrary HTML-like strings
4	+	fn html_fragment() -> impl Strategy<Value = String> {
5	+	let tags = prop::sample::select(vec![
6	+	"p", "div", "span", "strong", "em", "a", "h1", "h2", "h3",
7	+	"ul", "ol", "li", "blockquote", "pre", "code", "br", "hr",
8	+	"img", "table", "tr", "td", "th", "b", "i", "del", "sup", "sub",
9	+	]);
10	+
11	+	let text = "[a-zA-Z0-9 .,!?]{0,100}";
12	+
13	+	prop::collection::vec(
14	+	prop_oneof![
15	+	// Plain text
16	+	text.prop_map(\|s\| s),
17	+	// Opening + closing tag with text
18	+	(tags.clone(), text).prop_map(\|(tag, content)\| {
19	+	format!("<{tag}>{content}</{tag}>")
20	+	}),
21	+	// Self-closing tag
22	+	tags.clone().prop_map(\|tag\| format!("<{tag}/>")),
23	+	// Nested tags
24	+	(tags.clone(), tags.clone(), text).prop_map(\|(outer, inner, content)\| {
25	+	format!("<{outer}><{inner}>{content}</{inner}></{outer}>")
26	+	}),
27	+	],
28	+	1..10,
29	+	)
30	+	.prop_map(\|parts\| parts.join(""))
31	+	}
32	+
33	+	proptest! {
34	+	#[test]
35	+	fn never_panics(html in html_fragment()) {
36	+	let _ = pter::convert(&html);
37	+	}
38	+
39	+	#[test]
40	+	fn never_panics_on_arbitrary_bytes(s in "\\PC{0,500}") {
41	+	let _ = pter::convert(&s);
42	+	}
43	+
44	+	#[test]
45	+	fn output_contains_no_html_tags(html in html_fragment()) {
46	+	let md = pter::convert(&html);
47	+	// Output should never contain raw HTML tags
48	+	// (except inside code blocks, which we skip checking)
49	+	let without_code_blocks: String = md
50	+	.split("```")
51	+	.enumerate()
52	+	.filter(\|(i, _)\| i % 2 == 0) // only outside code blocks
53	+	.map(\|(_, s)\| s)
54	+	.collect();
55	+
56	+	// No <script>, <style>, <div>, etc. should leak through
57	+	assert!(!without_code_blocks.contains("<script"), "leaked <script> in: {md}");
58	+	assert!(!without_code_blocks.contains("<style"), "leaked <style> in: {md}");
59	+	assert!(!without_code_blocks.contains("<head"), "leaked <head> in: {md}");
60	+	}
61	+
62	+	#[test]
63	+	fn output_is_valid_utf8(html in html_fragment()) {
64	+	let md = pter::convert(&html);
65	+	// String type guarantees UTF-8, but verify no replacement chars snuck in
66	+	// from bad entity decoding
67	+	assert!(!md.contains('\u{FFFD}'), "replacement char in: {md}");
68	+	}
69	+
70	+	#[test]
71	+	fn no_excessive_blank_lines(html in html_fragment()) {
72	+	let md = pter::convert(&html);
73	+	assert!(!md.contains("\n\n\n"), "triple newline in output: {md}");
74	+	}
75	+
76	+	#[test]
77	+	fn no_trailing_whitespace_on_lines(html in html_fragment()) {
78	+	let md = pter::convert(&html);
79	+	for (i, line) in md.lines().enumerate() {
80	+	assert!(
81	+	line == line.trim_end(),
82	+	"trailing whitespace on line {i}: '{line}'"
83	+	);
84	+	}
85	+	}
86	+
87	+	#[test]
88	+	fn empty_input_returns_empty(s in "\\s{0,20}") {
89	+	let html = format!("<html><body>{s}</body></html>");
90	+	let md = pter::convert(&html);
91	+	// Whitespace-only input should produce empty or whitespace-only output
92	+	assert!(md.trim().is_empty() \|\| !s.trim().is_empty());
93	+	}
94	+	}