Querying richText field in GraphQL

squareclamp2 years ago

17

• 

  jmikrut

  2 years ago

  This is possible, but not in GraphQL unfortunately. Because GraphQL is so strict in its typing

  
  
  

  But in the REST API, here's how you'd do it

  
  
  

  ?where[content.children.text][like]=test

• 

  squareclamp2 years ago

  I see so, is it because the the type of content is JSON? If it was a string, would it work? Think if I could have a hidden field that is string for full text search. I'm not sure though what might be the best way to do full text search using Mongo.

• 

  jmikrut

  2 years ago

  100%

  
  
  

  very good call

  
  
  

  you should build that into a plugin actually

  
  
  

  it would be amazing

  
  
  

  Mongo does support full text search actually, but we have not yet wired it into Payload. That however will be making its way into core for sure

• 

  squareclamp2 years ago

  Thanks, I got something working using a hidden text field and

  beforeSave

  hook. I guess it could be packaged into a lightweight plugin.

  
  
  

  I'm doing this in the beforeSave hook for the collection -

  
  
  beforeChange: [ async ({ data, // incoming data to update or create with req, // full express request operation, // name of the operation ie. 'create', 'update' originalDoc, // original document }) => { data.indexableExcerpt = data.excerpt.flatMap( block => block.children.map( (b) => b.text)).join("") return data; // Return data to either create or update a document with } ]
  
  

  fields

  
  
  richText({ name: 'excerpt', }), { name: 'indexableExcerpt', type: 'text', index: true, admin: { hidden: true } }
  
  

  Is there a cleaner way to get rawText for richText field?

• 

  jmikrut

  2 years ago

  i think that's pretty nice tbh

  
  
  

  here's how we've done it before (the rich text list cell component)

  
  
  https://github.com/payloadcms/payload/blob/master/src/admin/components/views/collections/List/Cell/field-types/Richtext/index.tsx
• 

  squareclamp2 years ago

  Thanks, it seems like

  index: true

  would be kind of useless in this case. Since it probably just using a tree-based index. And we'd want to create text index and search operator for this use case which might be some work.

  https://www.digitalocean.com/community/tutorials/how-to-perform-full-text-search-in-mongodb
  
  

  I'll probably just go with something like Algolia for now. Will keep an eye out for full text index support or maybe contribute it 🙂

  
  
  

  IIUC, this would be scanning all documents even with index -

  
  
  { Posts(where: { indexableExcerpt: {contains: "small"} }, draft: true) { docs { id title excerpt } } }
• 

  bombnp2 years ago

  I actually did something similar to this. I created a function to recursively flatten all text elements, and put it to another hidden text field. Not sure if this can help.

  
  

  import { Descendant, Element, Text } from 'slate'
  
  export const flattenText = (obj: Descendant | Descendant[]): string => {
    if (Element.isElementList(obj) || Text.isTextList(obj)) {
      return obj.map(flattenText).join(' ')
    } else if (Element.isElement(obj)) {
      return flattenText(obj.children)
    } else if (Text.isText(obj)) {
      return obj.text
    } else {
      throw new Error('Invalid slate element:' + JSON.stringify(obj))
    }
  }

• 

  ssyberg2 years ago

  @495380410813906945

  even using agolia / 3rd party search, do you have an idea of how you'd integrate that in to Payload search?

• 

  squareclamp2 years ago

  I haven't tried Algolia. I've actually been wanting to try to create full text index manually or via Mongoose and verify if the index is used. If it is then perhaps we don't third part full text search provider.

  
  
  @364124941832159242

  Any suggestions on what might be a good place to create text index for a collection? I was think

  onInit

  in payload.init.

• 

  jmikrut

  2 years ago

  i would have a hook on the collection(s) that you want to index, and just keep the text index up to date whenever documents are created / updated

  
  
  

  not

  onInit
  
  

  that is how we do it for this site:

  
  
  https://hopenetwork.org
  
  

  take a look at the global search - it's all using Payload, and just uses

  afterChange

  hooks on pages, posts, etc.

• 

  squareclamp2 years ago

  Would it be possible to share snippet on how you might update index in the hook? I was expecting once you create index in

  onInit

  the index is updated automatically when fields change.

  
  
  

  The website search is looks great, and is very similar to my use case. One thing I recently realized was that adding

  highlights

  is an atlas only feature. So, you can only show documents but not highlight keywords.

  
  
  

  I actually really like the highlights (or underline) of the keyword when searching payload docs. Was curious if something like that might also be possible with MongoDB atlas.

• 

  jmikrut

  2 years ago

  that seems like it would be a frontend feature which should certainly be possible

  
  
  

  we will open-source our entire search index implementation for the Hope site

  
  
  

  that is a good call

• 

  squareclamp2 years ago

  awesome, thanks!

• 

  techinvertedlast year

  hey

  @364124941832159242

  , any update on open sourcing the search implementation

• 

  jmikrut

  last year

  hey! honestly i haven't thought about this in a minute but maybe we can expedite this for you

• 

  techinvertedlast year

  Hey

  @364124941832159242

  , would really appreciate that. Thanks

  
• 

  lucaceck9512 months ago

  Hi everyone

  
  

  any updates for searching words in slate richtext?