Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion platforms/pictique/api/src/database/entities/Post.ts
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,10 @@ export class Post {
@Column("text")
text!: string; // was content

@Column("simple-array", { nullable: true })
// simple-json (JSON.stringify/parse) is used instead of simple-array because
// base64 data URLs contain literal commas, which simple-array's naive
// comma-join/split encoding corrupts on read.
@Column("simple-json", { nullable: true })
images!: string[]; // was mediaUrls

@OneToMany(() => Comment, (comment: Comment) => comment.post)
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,145 @@
import { MigrationInterface, QueryRunner } from "typeorm";

/**
* Re-encode the `posts.images` column from the legacy `simple-array` format
* (naive comma-join) to `simple-json` (JSON.stringify/parse).
*
* Both column types compile down to a plain `text` column, so there is no
* schema change — only the stored payloads need converting. Without this, rows
* written under the old encoding would throw a JSON.parse error the first time
* the entity is read back after the decorator switches to `simple-json`.
*
* Base64 data URLs (`data:<mime>;base64,<payload>`) always contain a comma in
* their MIME prefix, so the old comma-join is ambiguous. We recover the
* original array by splitting only on commas that immediately precede a new
* `data:` URL — the `data:` token cannot occur inside a base64 payload (whose
* alphabet excludes `:`), so this boundary is unambiguous for image posts.
* Values that are already valid JSON arrays are left untouched (idempotent).
*/
export class ReencodePostImages1784133148233 implements MigrationInterface {

// Number of rows loaded per keyset-paginated batch. Kept modest because
// each image payload can be a multi-MB base64 data URL.
private static readonly BATCH_SIZE = 200;

public async up(queryRunner: QueryRunner): Promise<void> {
const batchSize = ReencodePostImages1784133148233.BATCH_SIZE;
// Keyset cursor: the all-zero UUID is the lower bound, so `id > cursor`
// starts from the first row. Paging by id (never by the mutated
// `images` column) means re-encoded rows can't reappear in a later
// batch, so the walk always terminates.
let cursor = "00000000-0000-0000-0000-000000000000";
let batch: Array<{ id: string; images: string | null }>;

do {
batch = await queryRunner.query(
`SELECT "id", "images" FROM "posts"
WHERE "images" IS NOT NULL AND "id" > $1
ORDER BY "id" ASC
LIMIT $2`,
[cursor, batchSize],
);
if (batch.length === 0) break;
cursor = batch[batch.length - 1].id;

const updates: Array<{ id: string; images: string }> = [];
for (const row of batch) {
if (typeof row.images !== "string") continue;
const reencoded = this.reencode(row.images);
// `undefined` = already valid JSON, no write needed.
if (reencoded !== undefined) {
updates.push({ id: row.id, images: reencoded });
}
}

if (updates.length > 0) {
// Single bulk UPDATE per batch via a VALUES join, instead of
// one round-trip per row.
const valuesSql = updates
.map((_, i) => `($${i * 2 + 1}::uuid, $${i * 2 + 2}::text)`)
.join(", ");
const params = updates.flatMap((u) => [u.id, u.images]);
await queryRunner.query(
`UPDATE "posts" AS p
SET "images" = v.images
FROM (VALUES ${valuesSql}) AS v(id, images)
WHERE p."id" = v.id`,
params,
);
}
} while (batch.length === batchSize);
}

/**
* Convert a single legacy `images` value to its `simple-json` encoding.
* Returns the new string to store, or `undefined` when the value is already
* a valid JSON array and should be left untouched.
*/
private reencode(raw: string): string | undefined {
// Empty string is how simple-array encoded an empty array. Left as "",
// simple-json would choke on JSON.parse("") — convert to "[]".
if (raw === "") return "[]";

// Already migrated (valid JSON array) — leave as-is.
if (raw.trimStart().startsWith("[")) {
try {
JSON.parse(raw);
return undefined;
} catch {
// Not actually valid JSON; fall through and re-encode.
}
}

// Reconstruct the array from the legacy comma-joined string.
// Walk comma-separated tokens: a base64 data URL got split across two
// tokens ("data:<mime>;base64" + "<payload>") by the comma in its own
// prefix, so rejoin that pair. Any other value (e.g. a Firebase/HTTP
// download URL) contains no internal comma and stands alone. This
// recovers pure-data, pure-URL, and mixed posts in any order.
const tokens = raw.split(",");
const images: string[] = [];
for (let i = 0; i < tokens.length; i++) {
const token = tokens[i].trim();
if (token === "") continue;

if (token.startsWith("data:")) {
const payload = (tokens[i + 1] ?? "").trim();
images.push(payload ? `${token},${payload}` : token);
i++; // consume the payload token
} else {
images.push(token);
}
}
Comment thread
coderabbitai[bot] marked this conversation as resolved.

return JSON.stringify(images);
}

public async down(queryRunner: QueryRunner): Promise<void> {
// Reverse: re-encode JSON arrays back to the legacy comma-joined format.
const rows: Array<{ id: string; images: string | null }> =
await queryRunner.query(
`SELECT "id", "images" FROM "posts" WHERE "images" IS NOT NULL AND "images" <> ''`,
);

for (const row of rows) {
const raw = row.images;
if (raw === null || raw === "") continue;

let images: string[];
try {
const parsed = JSON.parse(raw);
if (!Array.isArray(parsed)) continue;
images = parsed;
} catch {
// Not JSON — already in legacy format.
continue;
}

await queryRunner.query(
`UPDATE "posts" SET "images" = $1 WHERE "id" = $2`,
[images.join(","), row.id],
);
}
}

}
28 changes: 1 addition & 27 deletions platforms/pictique/client/src/lib/fragments/Post/Post.svelte
Original file line number Diff line number Diff line change
Expand Up @@ -29,36 +29,11 @@
options?: Array<{ name: string; handler: () => void }>;
}

function pairAndJoinChunks(chunks: string[]): string[] {
const result: string[] = [];

console.log('chunks', chunks);
for (let i = 0; i < chunks.length; i += 2) {
const dataPart = chunks[i];
const chunkPart = chunks[i + 1];

if (dataPart && chunkPart) {
if (dataPart.startsWith('data:')) {
result.push(`${dataPart},${chunkPart}`);
} else {
result.push(dataPart);
result.push(chunkPart);
}
} else {
if (!dataPart.startsWith('data:')) result.push(dataPart);
console.warn(`Skipping incomplete pair at index ${i}`);
}
}
console.log('result', result);

return result;
}

const {
avatar,
userId,
username,
imgUris: uris,
imgUris,
text,
count,
callback,
Expand All @@ -67,7 +42,6 @@
...restProps
}: IPostProps = $props();

let imgUris = $derived(pairAndJoinChunks(uris));
let galleryRef: HTMLDivElement | undefined = $state();
let currentIndex = $state(0);

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -42,36 +42,11 @@
ownerProfile?: userProfile;
}

function pairAndJoinChunks(chunks: string[]): string[] {
const result: string[] = [];

console.log('chunks', chunks);
for (let i = 0; i < chunks.length; i += 2) {
const dataPart = chunks[i];
const chunkPart = chunks[i + 1];

if (dataPart && chunkPart) {
if (dataPart.startsWith('data:')) {
result.push(`${dataPart},${chunkPart}`);
} else {
result.push(dataPart);
result.push(chunkPart);
}
} else {
if (!dataPart.startsWith('data:')) result.push(dataPart);
console.warn(`Skipping incomplete pair at index ${i}`);
}
}
console.log('result', result);

return result;
}

const {
avatar,
userId,
username,
imgUris: uris,
imgUris,
text,
count,
callback,
Expand All @@ -82,7 +57,6 @@
...restProps
}: IPostProps = $props();

let imgUris = $derived(pairAndJoinChunks(uris));
let galleryRef: HTMLDivElement | undefined = $state();
let currentIndex = $state(0);
let commentValue = $state('');
Expand Down
Loading