[BUG]: Nested (atomic) template tags break rendering on special inputs

[jaxley]

Was having this issue with 2.4.0beta6 and beta7. Downgraded to 2.3.1 and still has this issue

Readwise Mirror: Error during sync: Template render error: (unknown path) [Line 32, Column 25]
  unexpected token: }
    at Object._prettifyError (plugin:readwise-mirror:466:13)
    at Template.render (plugin:readwise-mirror:3062:23)
    at Environment.renderString (plugin:readwise-mirror:2933:19)
    at Atomizer.atomize (plugin:readwise-mirror:7377:36)
    at ReadwiseMirror.processReadwiseLibrary (plugin:readwise-mirror:17361:47)
    at async ReadwiseMirror.writeLibraryToMarkdown (plugin:readwise-mirror:17207:31)
    at async ReadwiseMirror.sync (plugin:readwise-mirror:17458:17)
    at async ReadwiseMirror.download (plugin:readwise-mirror:17490:9)
error @ plugin:readwise-mirror:8863
sync @ plugin:readwise-mirror:17477
await in sync
download @ plugin:readwise-mirror:17490
await in download
K6 @ app.js:1
t.onChooseItem @ app.js:1
t.onChooseSuggestion @ app.js:1
t.selectSuggestion @ app.js:1
e.useSelectedItem @ app.js:1
(anonymous) @ app.js:1
e.handleKey @ app.js:1
e.onKeyEvent @ app.js:1

It seems to fail on a specific book, processing a specific deleted highlight.

{
    "id": 899850250,
    "is_deleted": true,
    "text": "At this juncture, the \"we didn't know\" or \"we don't have the expertise\" excuses do not fly anymore. If you don't know how to balance your books and comply with tax laws and regulations, you hire an accountant. If you don't know anything about labor regulations, running payroll or managing personnel, you get yourself a human resources professional. If you don't know *anything* about *any* law at all, you hire a lawyer. Don't know anything about cybersecurity? You know what to do. That's the price of doing business right now.",
    "location": 2119011,
    "location_type": "offset",
    "note": "",
    "color": "",
    "highlighted_at": "2025-06-06T16:58:47.520Z",
    "created_at": "2025-06-06T16:58:47.575Z",
    "updated_at": "2025-06-06T16:58:47.575Z",
    "external_id": "01jx32wha0znpgkt1ceb0jk38n",
    "end_location": null,
    "url": "https://read.readwise.io/read/01jx32wha0znpgkt1ceb0jk38n",
    "book_id": 52230589,
    "tags": [],
    "is_favorite": false,
    "is_discard": false,
    "readwise_url": "https://readwise.io/open/899850250"
}

[johannrichard]

Thanks for the report and example data, most useful. Will have a look at it.

[johannrichard]

Also: Would you mind posting your highlight template?

[jaxley]

Probably the biggest architectural issue here is that whatever exception occurs, it stops ALL synchronizing - so you can never get past it. Shouldn't be a fatal error except for the problematic entry (would be helpful to include data on which record was being processed in the exception as well)

I did some additional debugging. Didn't find the smoking gun yet (this particular issue is annoying as it seems to be from a callback or other async call so the full call stack is not available when the exception occurs). Took a bit to get breakpoints in the right spots to at least get more data when this happens.

The previous info may be a red herring (those were the last messages to the console before the exception - however, they were consistent.

When the exception occurs, this is what ctx is (from renderString IIRC):

{
    "basename": "update-only-deletions-beetbox-beets-discussion-5461",
    "doc": {
        "id": 52642233,
        "readwise_url": "https://readwise.io/bookreview/52642233",
        "unique_url": "https://read.readwise.io/read/01jy7mm3k49d4fvrvebkavx8c4",
        "source_url": "https://github.com/beetbox/beets/discussions/5461",
        "title": "Update Only Deletions? · Beetbox/Beets · Discussion #5461",
        "sanitized_title": "update-only-deletions-beetbox-beets-discussion-5461",
        "author": [
            "https://github.com/beetbox/"
        ],
        "authorStr": "[[https://github.com/beetbox/]]",
        "document_note": null,
        "summary": "The user wants the update command to only detect deleted files without checking others. The maintainer says this feature doesn't exist yet but could be added easily with a flag. Meanwhile, a shell script or the beet update -F deleted command can help identify deleted files.",
        "category": "articles",
        "num_highlights": 1,
        "created": "2025-06-20",
        "updated": "2025-06-20",
        "cover_image_url": "https://opengraph.githubassets.com/fe783b4d0cad6e54ac436a0b2daa2faf218fb66527f6fb3f53d9cb15ef6dcb10/beetbox/beets/discussions/5461",
        "highlights": [
            {
                "id": 905211544,
                "is_deleted": false,
                "text": "As an alternative, you could use a shell script like this to only run update on deleted items:\n\n#!/usr/bin/env bash\n\nQUERY=()\nwhile read -r LINE; do\n\tFILE_ID=\"${LINE%%$'\\t'*}\"\n\tFILE_PATH=\"${LINE#*$'\\t'}\"\n\tif ! [ -f \"$FILE_PATH\" ]; then\n\t\t# Add deleted items to query.\n\t\tQUERY+=(\"id:$FILE_ID\" \",\")\n\tfi\ndone < <(beet ls -f $'$id\\t$path')\nif [ ${#QUERY[@]} -gt 0 ]; then\n\tunset \"QUERY[-1]\" # Remove trailing comma.\n\tbeet -vvv update --pretend \"${QUERY[@]}\"\nelse\n\techo \"Nothing to do.\"\nfi",
                "location": 56143,
                "location_type": "offset",
                "note": "",
                "color": "",
                "highlighted_at": "2025-06-20T21:41:33.276Z",
                "created_at": "2025-06-20T21:41:33.324Z",
                "updated_at": "2025-06-20T21:41:33.324Z",
                "external_id": "01jy7mmawr2p6wjercmn2vyvbm",
                "end_location": null,
                "url": "https://read.readwise.io/read/01jy7mmawr2p6wjercmn2vyvbm",
                "book_id": 52642233,
                "tags": [],
                "is_favorite": false,
                "is_discard": false,
                "readwise_url": "https://readwise.io/open/905211544"
            }
        ],
        "last_highlight_at": "2025-06-20",
        "tags": "",
        "highlight_tags": "",
        "tags_nohash": "",
        "hl_tags_nohash": "",
        "linktext": "update-only-deletions-beetbox-beets-discussion-5461"
    },
    "book": {
        "user_book_id": 52642233,
        "is_deleted": false,
        "title": "Update Only Deletions? · Beetbox/Beets · Discussion #5461",
        "author": "https://github.com/beetbox/",
        "readable_title": "Update Only Deletions? · Beetbox/Beets · Discussion #5461",
        "source": "reader",
        "cover_image_url": "https://opengraph.githubassets.com/fe783b4d0cad6e54ac436a0b2daa2faf218fb66527f6fb3f53d9cb15ef6dcb10/beetbox/beets/discussions/5461",
        "unique_url": "https://read.readwise.io/read/01jy7mm3k49d4fvrvebkavx8c4",
        "summary": "The user wants the update command to only detect deleted files without checking others. The maintainer says this feature doesn't exist yet but could be added easily with a flag. Meanwhile, a shell script or the beet update -F deleted command can help identify deleted files.",
        "book_tags": [],
        "category": "articles",
        "document_note": null,
        "readwise_url": "https://readwise.io/bookreview/52642233",
        "source_url": "https://github.com/beetbox/beets/discussions/5461",
        "external_id": "01jy7mm3k49d4fvrvebkavx8c4",
        "asin": null,
        "highlights": [
            {
                "id": 905211544,
                "is_deleted": false,
                "text": "As an alternative, you could use a shell script like this to only run update on deleted items:\n\n#!/usr/bin/env bash\n\nQUERY=()\nwhile read -r LINE; do\n\tFILE_ID=\"${LINE%%$'\\t'*}\"\n\tFILE_PATH=\"${LINE#*$'\\t'}\"\n\tif ! [ -f \"$FILE_PATH\" ]; then\n\t\t# Add deleted items to query.\n\t\tQUERY+=(\"id:$FILE_ID\" \",\")\n\tfi\ndone < <(beet ls -f $'$id\\t$path')\nif [ ${#QUERY[@]} -gt 0 ]; then\n\tunset \"QUERY[-1]\" # Remove trailing comma.\n\tbeet -vvv update --pretend \"${QUERY[@]}\"\nelse\n\techo \"Nothing to do.\"\nfi",
                "location": 56143,
                "location_type": "offset",
                "note": "",
                "color": "",
                "highlighted_at": "2025-06-20T21:41:33.276Z",
                "created_at": "2025-06-20T21:41:33.324Z",
                "updated_at": "2025-06-20T21:41:33.324Z",
                "external_id": "01jy7mmawr2p6wjercmn2vyvbm",
                "end_location": null,
                "url": "https://read.readwise.io/read/01jy7mmawr2p6wjercmn2vyvbm",
                "book_id": 52642233,
                "tags": [],
                "is_favorite": false,
                "is_discard": false,
                "readwise_url": "https://readwise.io/open/905211544"
            }
        ]
    }
}

tmplStr is:

%%
ID:: 52642233
Updated:: 2025-06-20
%%

![](https://opengraph.githubassets.com/fe783b4d0cad6e54ac436a0b2daa2faf218fb66527f6fb3f53d9cb15ef6dcb10/beetbox/beets/discussions/5461)

# Update Only Deletions? · Beetbox/Beets · Discussion #5461

## About
Title:: [[Update Only Deletions? · Beetbox/Beets · Discussion #5461]]
Authors:: [[https://github.com/beetbox/]]
Category:: #articles
NumberOfHighlights:: ==1==
ReadwiseURL:: https://readwise.io/bookreview/52642233
SourceURL: https://github.com/beetbox/beets/discussions/5461
UpdatedDate:: [[2025-06-20]]
LastHighlighted:: *2025-06-20*
---

## Highlights


  As an alternative, you could use a shell script like this to only run update on deleted items:

#!/usr/bin/env bash

QUERY=()
while read -r LINE; do
	FILE_ID="${LINE%%$'\t'*}"
	FILE_PATH="${LINE#*$'\t'}"
	if ! [ -f "$FILE_PATH" ]; then
		# Add deleted items to query.
		QUERY+=("id:$FILE_ID" ",")
	fi
done < <(beet ls -f $'$id\t$path')
if [ ${#QUERY[@]} -gt 0 ]; then
	unset "QUERY[-1]" # Remove trailing comma.
	beet -vvv update --pretend "${QUERY[@]}"
else
	echo "Nothing to do."
fi ^905211544

[jaxley]

BTW, I tried using replace("}", "rightcurly") | replace("{", "leftcurly") in the template to see if it was an issue with the curly braces in the code snippet. That didn't change the exception.

[johannrichard]

Found the "culprit": It's not the curly braces, but the %% in the following line:

FILE_ID="${LINE%%$'\t'*}"

I'll have to investigate but It's possibly related to the atomizer / atomic parsing. Might take some time, I'm afraid, need to figure out what's different (Unfortunately right now even if Atomizer is turned off we'll be hitting that error).

[jaxley]

Okay, that makes sense. I was starting to get a feel for the code flow and suspected something there. I'm going to see about building some unit tests to repro the bug directly in atomizer. Jason

[jaxley]

Good news! I created a unit test that reproduces the issue. O don't have a solution to that parsing issue but I was able to fox the code so a single article does not break sync of all other files. Will put a PR together Wednesday. I found a couple of other errors that shouldn't be fatal that break sync. Will see about fixing those too. Jason

…

-------- Original Message --------

On Tuesday, 06/02/26 at 14:24 Jason Axley ***@***.***> wrote: Okay, that makes sense. I was starting to get a feel for the code flow and suspected something there. I'm going to see about building some unit tests to repro the bug directly in atomizer. Jason

[johannrichard]

Thanks @jaxley for your patience and contribution. The workaround is released in 2.4.0, will hopefully soon find time to work on a fix.