-
-
Notifications
You must be signed in to change notification settings - Fork 1.3k
Returning Invalid syntax for type json on arrays #442
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
We convert javascript arrays into postgres arrays. If you stringify your input array it should work: client.query('INSERT INTO my_table (data) VALUES ($1)', [ JSON.stringify([1,2,3])], handler); We don't now the column type in advance so we can't do a auto conversions of the array. |
@booo Yeah, I was thinking that may be the case. Well, in that case, non-issue! |
@booo @brianc This behavior is a bug. Converting data from [{}, {}] to {{}, {}} should not be determined by pg. The only solution is to use JSON.stringify before running the insert query? What if you're running an ORM? What if you are unaware of this limitation? Just because pg doesn't know the column type in advance does not mean you should convert to the syntax of column type "array". It might be helpful, but it's also a pretty big assumption and makes it really hard to use an array in a json field, despite it being supported and normal behavior by the database. The better behavior would be to leave converting data types to the client. Both possibilities would be easy. As is, it's really hard to get the expected behavior using what's available. |
see also: #374 |
@badave Doesn't your proposed behavior require the same amount of work to insert JSON? You're gonna be stringifying the JSON anyway. |
Proposed behavior passing {} or [] into a column without the library being opinionated about converting it into a postgres array type. The larger problem isn't the JSON.stringify, it's the behavior of pg. As a matter of fact, the point should be that I shouldn't have to do a JSON.stringify because the library already handles it with JSON objects without stringifying. This shouldn't be a mandatory conversion. It's that simple. The database error is "invalid input syntax for type json", an insufficient error message as a user to conclude its caused by the lib transforming the data. It's transforming valid JSON into invalid JSON. However, if I were using an array type and inserting an array into postgres with the incorrect format, I would understand the invalid input syntax for type array because it isn't JSON format, it requires an array format. At that point I would look for a solution that converted my data into the correct form to insert into the pg. And I could understand it better if the array column type was any good in postgres, but it's complete crap. I'd prefer to use a json array because it is more recognizable. |
Yeah, I see what you're saying. It's a bit funky that node-pg assumes to stringify Objects but not Arrays. I think making it more consistent, or rather, no incoming magic (I like the magic coming out though!) would be for the best. It seems a little arbitrary, especially now that JSON has been added to the data-type mix, that JS arrays are assumed to be PG Arrays, when they could be type json or even hstore. My preference would be to make no assumptions and leave it up to the library consumer. I'm not sure if that means node-pg should JSON.stringify Arrays and Objects by default, as that would be trading one magic for another. But it certainly does seem like a sane default. |
Hi, this issue should be reopened, since in postgres shell is possible to execute:
Best Regards. |
I'm still running into this error even after stringifying the input array. Here's my JSON object: [
{ serviceId: 2, petId: 5, checked: true },
{ serviceId: 3, petId: null, checked: false }
] After stringifying: "[{"serviceId":2,"petId":5,"checked":true},{"serviceId":3,"petId":null,"checked":false}]" I insert and receive the following error:
It appears the input is being stringified again. Am I missing something here? Appreciate any help! |
Here's an example of using a json data type & doing a round-trip query from node -> postgres -> and back to node with the json type preserved: Hope that helps! |
@brianc thanks for this. However, isn't this just testing on a single object and not an array of objects? |
@jamesdixon using pg for arrays of objects is working fine for me: require('pg').connect( (error, client, done) => {
if (error) throw error
client.query('select $1::json as arr', [JSON.stringify([{foo: 'bar'}])], (error, result) => {
done()
if (error) throw error
console.log(result.rows[0].arr) // => [{ foo: 'bar' }]
process.exit(0)
})
}) |
If you stringify first, you'll need to cast the value to |
@jrf0110 thanks for this. I've confirmed this does work on insert as well. It appears something is double-stringifying my input. Appreciate the help, fellas! |
@charmander so are issue 442 and 1143 overlapping issues? I see they are going back between open/close. @joskuijpers seemed to find my issue, I was wondering if any progress had been made regarding a solution |
@PythonDevOp Yes, they’re the same issue. I’m not sure how @brianc wants to solve it, if at all, but you can use |
If you need to pass in an array of objects as a single JSON value then do what @jrf0110 has done and call |
Would it make sense for |
A way to configure the default behavior would be desirable. |
Another option is to make your column have a type of |
I found the file @brianc referred to, since the link is now broken: https://github.com/brianc/node-postgres/blob/9274f08fa2d8ae55a218255bf7880d26b6abc935/test/integration/client/json-type-parsing-tests.js |
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
This is a **BREAKING** change that: - adds tests against the [upstream `sharedb` DB test suite][1] - adds a CI build for the tests against current Node.js and Postgres versions - breaks the API to conform to the upstream tests, including adding metadata support The breaks are: - Dropping non-null constraints on `snapshots.doc_type` and `snapshots.data` (to allow `Doc`s to be deleted) - Adding a new `snapshots.metadata` `json` column - Respecting `options.metadata` and `fields.$submit`, which were previously ignored on `getOps()`, and useless on `getSnapshot()` (which didn't store metadata) - `snapshot.m` is now `undefined` if not present, or `null` if unrequested (inline with the spec) On top of this it also makes some bugfixes to conform to the spec: - Ignore unique key validations when committing, since this may happen during concurrent commits - `JSON.stringify()` JSON fields, which [break][2] if passed a raw array - Default `from = 0` if unset in `getOps()` [1]: https://github.com/share/sharedb/blob/7abe65049add9b58e1df638aa34e7ca2c0a1fcfa/test/db.js#L25 [2]: brianc/node-postgres#442
12 years later I'm stuck with this. I use pgtyped to execute my sql. I try stringify() before passing to pgtyped's generated function. No difference. I've tried different combinations of cast and quotes in the sql without luck. I'm not going to persevere. My arrays that were inserting fine when they were buried in an object cannot be the root of a json column value with my current setup, it seems. |
open to a PR if you find a way to fix this! |
@bbsimonbb Both of those approaches do work, so if you’d like some help, post the usual necessities (minimal reproducible example, debugging details). |
When your column is type json and you try and insert a JSON array, you get the following error:
For example, this works in PSQL:
But this returns an error in node-pg:
The text was updated successfully, but these errors were encountered: