[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16754413#comment-16754413 ] David Smiley commented on SOLR-7672: How so [~mkhludnev]? I _wish_ it were so. The recent improvements to nested docs in Solr have yet to touch the block join queries. In 8.0, I think these query parsers could assume {{parent:-\_nest_path_:*}} but only if this field is defined. This approach is taken with the updated child doc transformer. > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev >Priority: Major > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15729289#comment-15729289 ] Romain Rigaux commented on SOLR-7672: - e.g. if we take the book/review example of http://yonik.com/solr-nested-objects/ This would (field name, then hierarchy of values) let us query to introspect and build the "schema" easily: {code} _type_:"type_s:book" _type_:"type_s:book.review" {code} > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev > Fix For: 5.5, 6.0 > > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15632831#comment-15632831 ] Romain Rigaux commented on SOLR-7672: - +1 Having the "paths/types" and name of the "type" fields would provide out of a the box a "schema" of the data that would greatly improve the user experience with nested objects. > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev > Fix For: 5.5, 6.0 > > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15038194#comment-15038194 ] Mikhail Khludnev commented on SOLR-7672: just an idea for further consideration, when {{q=\{!parent}foo=\*}}, it may imply {{fl=\[child],*}} just to show children by default. > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev > Fix For: 5.4, Trunk > > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15039542#comment-15039542 ] Yonik Seeley commented on SOLR-7672: Yeah, that probably works. We can always just use some syntactic sugar if we don't want a literal \_root\_ mandated in whatever future query syntax we come up with. Note to new readers: when you see italicized text, we normally don't mean it that way... it's JIRA markup. Normally what we really mean is a literal leading and trailing underscore on whatever is being proposed. {code}_type_:_root_ is what is being proposed here.{code} And if you see backslashes in front of the underscores, you're probably reading the post in raw form (i.e. from an email) and it means that the author is just trying to get JIRA to display literal underscores (but not backslashes). > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev > Fix For: 5.4, Trunk > > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15038213#comment-15038213 ] Mikhail Khludnev commented on SOLR-7672: [~ysee...@gmail.com] my turn: {{_type_:_root_}} and spin off a jira about {{_type_:_root_.home_address}} or {{_type_:home_address}} Do you like it? > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev > Fix For: 5.4, Trunk > > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (SOLR-7672) introduce implicit _parent_:true
[ https://issues.apache.org/jira/browse/SOLR-7672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15027313#comment-15027313 ] Yonik Seeley commented on SOLR-7672: Hmmm, \_parent\_:true wouldn't seem to work that well for multi-level hierarchies (remember, we're not limited to a single parent-child level). Seems like something more general is desirable. It feels like indexing \_path\_ would be both general and very useful. The top-level parents would have a zero length _path_:"" Unfortunately, "/" is used by regex in the lucene query parser, so we could use "." as a path separator instead. This fits in well with (future) auto-mapping from JSON structure to nested documents. Example: {code} { id:x, work_address: { state:NY, street:"my work street" }, home_address:{ state:NJ, street:"my_home street" }, cars:[ { make: Toyota, model: Highlander }, { make: Honda, model: Accord } ] } {code} _path_ would be "" for the root, then "home_address", "street_address", and "cars" for the other nested documents. Note: I'm not asking that all this be *done* in this issue, just that we think about the end-game for nested documents (where we want to go), and make whatever we do fit in with that scheme. For this specific JIRA issue, all that would need to be done is use a more generic \_path\_:"" Hmmm, perhaps \_type\_ would be a better name, since we aren't including id values or anything, we're just deriving a type name from the path. > introduce implicit _parent_:true > -- > > Key: SOLR-7672 > URL: https://issues.apache.org/jira/browse/SOLR-7672 > Project: Solr > Issue Type: Improvement > Components: query parsers, update >Affects Versions: 5.2 >Reporter: Mikhail Khludnev >Assignee: Mikhail Khludnev > Fix For: 5.4, Trunk > > > Solr provides block join support in non-invasive manner. It turns out, it > gives a chance to shoot a leg. As it was advised by [~thetaphi] at SOLR-7606, > let AddUpdateCommand add _parent_:true field to the document (not to > children). Do it *always* no matter whether it has children or not. > Also, introduce default values for for block join qparsers \{!parent > *which=\_parent\_:true*} \{!child *of=\_parent\_:true*} (sometimes, I rather > want to hide them from the user, because they are misunderstood quite often). > > Please share your concerns and vote. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org