[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16875631#comment-16875631 ] Paul Rogers edited comment on DRILL-7308 at 6/30/19 1:55 AM: - Modified the {{SchemaBuilder}} class to do exactly what I said we don't want to do: it avoids setting the precision if the precision is zero. This allows the (wrong) code in the REST feature to work. Still, the incorrect code should change as explained above to avoid breaking the next time someone sets a precision of 0. Also removed the empty schema batch so that simple queries return just one batch of data. The result is that the broken code in the REST call should work for simple one-batch queries. Nothing I can do, however, will fix the fact that the schema will be repeated for every batch; fixing that will require changes to the REST code itself. was (Author: paul.rogers): Modified the {{SchemaBuilder}} class to do exactly what I said we don't want to do: it avoids setting the precision if the precision is zero. This allows the (wrong) code in this feature to work. The incorrect code should change. Also removed the empty schema batch so that simple queries return just one batch of data. The result is that the broken code in the REST call should work for simple one-batch queries. Nothing I can do, however, will fix the fact that the schema will be repeated for every batch; fixing that will require changes to the REST code itself. > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following metadata: > {code:sql} > SELECT * FROM dfs.test.`domains.csvh` LIMIT 1 > {code} > {code:json} > { > "queryId": "22eee85f-c02c-5878-9735-091d18788061", > "columns": [ > "domain" > ], > "rows": [} > { "domain": "thedataist.com" } ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > "queryState": "COMPLETED", > "attemptedAutoLimit": 0 > } > {code} > There are two issues here: > 1. VARCHAR now has precision > 2. There are twice as many columns as there should be. > Additionally, if you query a regular CSV, without the columns extracted, you > get the following: > {code:json} > "rows": [ > { > "columns": "[\"ACCT_NUM\",\"PRODUCT\",\"MONTH\",\"REVENUE\"]" } > ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > {code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16872033#comment-16872033 ] Paul Rogers edited comment on DRILL-7308 at 6/30/19 1:54 AM: - Recall that Drill can return not only multiple batches, but multiple "result sets": runs of batches with different schemas. A more sophisticated REST solution would handle this case. I can't find any ProtoBuf field that says that the schema changed. Instead, we'd have to reuse code from elsewhere which compares the current schema to the previous one. Ideally, in that case, we'd create a new JSON element for the second schema. Something like: {code:json} { resultSets: [ { "rows": ... "schema": ... }, { "rows": ... "schema": ... } ] } {code} It is easy to create such a case. Simply create two CSV files, one with 2 columns, the other with three. Use just a simple \{{SELECT * FROM yourTable}} query. You will get two data batches, each with a distinct schema. The current implementation will give just the first schema and all rows, with varying schemas. (Actually, the current implementation will list the two columns, then the three columns, duplicating the first two, but we want to fix that...) This is yet another reason to use a provisioned schema: with such a schema we can guarantee that the entire query will return a single, consistent schema regardless of the variation across files. A quick & dirty solution is to clear and rebuild the schema objects on every batch. That way, the value sent to the user will reflect the last schema which, if you are lucky, will be valid for the initial batches as well as later batches. It is a known open, unresolved issue that Drill does not attempt to merge schema changes, and that unmerged schema changes cannot be handled by ODBC or JDBC clients. We can assume, however, that the users of the REST API won't have messy data and won't run into this issue. was (Author: paul.rogers): Recall that Drill can return not only multiple batches, but multiple "result sets": runs of batches with different schemas. A more sophisticated REST solution would handle this case. I can't find any ProtoBuf field that says that the schema changed. Instead, we'd have to reuse code from elsewhere which compares the current schema to the previous one. Ideally, in that case, we'd create a new JSON element for the second schema. Something like: {code:json} { resultSets: [ { "rows": ... "schema": ... }, { "rows": ... "schema": ... } ] } {code} It is easy to create such a case. Simply create two CSV files, one with 2 columns, the other with three. Use just a simple \{{SELECT * FROM yourTable}} query. You will get two data batches, each with a distinct schema. The current implementation will give just the first schema and all rows, with varying schemas. (Actually, the current implementation will list the two columns, then the three columns, duplicating the first two, but we want to fix that...) This is yet another reason to use a provisioned schema: with such a schema we can guarantee that the entire query will return a single, consistent schema regardless of the variation across files. > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following metadata: > {code:sql} > SELECT * FROM dfs.test.`domains.csvh` LIMIT 1 > {code} > {code:json} > { > "queryId": "22eee85f-c02c-5878-9735-091d18788061", > "columns": [ > "domain" > ], > "rows": [} > { "domain": "thedataist.com" } ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > "queryState": "COMPLETED", > "attemptedAutoLimit": 0 > } > {code} > There are two issues here: > 1. VARCHAR now has precision > 2. There are twice as many columns as there should be. > Additionally, if you query a regular CSV, without the columns extracted, you > get the following: > {code:json} > "rows": [ > { > "columns": "[\"ACCT_NUM\",\"PRODUCT\",\"MONTH\",\"REVENUE\"]" } > ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > {code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16872964#comment-16872964 ] Paul Rogers edited comment on DRILL-7308 at 6/29/19 6:13 PM: - [~cgivre], the problem here is that the code shown earlier is counting on a Protobuf implementation detail that is not actually a part of the Drill schema specification (to the degree there is such a specification.) For VarChar, a precision of 0 means that the user requested {{VARCHAR}}, while a precision of, say, 10 means the user requested {{VARCHAR(10)}}. The scale field is never valid for {{VARCHAR}}. The output of {{VARCHAR(0,0)}} is not a problem with the code that generated the schema. Instead, it is a problem with the way that the REST code attempts to generate a type name from the schema structures. To be more precise, the REST code incorrectly assumes that the {{isSet()}} methods are the correct way to check for a 0 value. This is an incorrect assumption. The Protobuf issue is that, unlike a regular Java object, if we never actually write to the precision field, then the value is unset. If we write, even if we write 0, the value is set. We certainly don't want to litter our code with things like: {code:java} if (precision != 0) { schemaBuilder.setPrecision(precision); } {code} So, the code that uses the schema objects should do the following to determine if the value is other than the default: both ask if the value is set, and if so, ask if the value is non-zero. As it turns out, the unset value is 0, so there is actually no need to ask if the value is set in this case. Taking a step back, the type formatting code should not even be in the REST API. The proper place for it is in {{Types}}. In fact, {{Types}} already has the desired function: {{getExtendedSqlTypeName()}}. However, this function only formats decimals; we need to add a case clause for VARCHAR. Note that {{getExtendedSqlTypeName()}} exposes the *SQL name* for types. The current REST implementation exposes the internal Drill name. That is, {{getExtendedSqlTypeName()}} will report, say, {{DOUBLE}} while the REST code will report {{Float8}}. This is probably a bug since the documentation explains the SQL types, not the internal types. That said, I actually have not seen any places in Drill where we set or use the VARCHAR width. So, no point in trying to format it. In this case, you can just use {{getExtendedSqlTypeName()}} directly as-is. Or, if we want to display the width, add the required code to that function. Please file a separate JIRA for the UDF issue. Please provide an attachment or link to a sample UDF. I'll see if I can track down that CSV-specific issue in case it relates to the EVF. was (Author: paul.rogers): [~cgivre], the problem here is that the code shown earlier is counting on a Protobuf implementation detail that is not actually a part of the Drill schema specification (to the degree there is such a specification.) For VarChar, a precision of 0 means that the user requested {{VARCHAR}}, while a precision of, say, 10 means the user requested {{VARCHAR(10}}. The scale is never valid for {{VARCHAR}}, it is an artifact of the incorrect way the above code was written. The Protobuf issue is that, unlike a regular Java object, if we never actually write to the precision field, then the value is unset. If we write, even if we write 0, the value is set. We certainly don't want to litter our code with things like: {code:java} if (precision != 0) { schemaBuilder.setPrecision(precision); } {code} So, we should ask if the precision is set and non-zero. In fact, the type formatting code should not even be in the REST API. The proper place for it is in {{Types}}. In fact, that class already has the desired function: {{getExtendedSqlTypeName()}}. However, this function only formats decimals; we need to add a case clause for VARCHAR. That said, I actually have not seen any places in Drill where we set or use the VARCHAR width. So, no point in trying to format it. In this case, you can just use {{getExtendedSqlTypeName()}} directly as-is. Please file a separate JIRA for the UDF issue. Please provide an attachment or link to a sample UDF. I'll see if I can track down that CSV-specific issue in case it relates to the EVF. > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following
[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16872964#comment-16872964 ] Paul Rogers edited comment on DRILL-7308 at 6/26/19 6:11 AM: - [~cgivre], the problem here is that the code shown earlier is counting on a Protobuf implementation detail that is not actually a part of the Drill schema specification (to the degree there is such a specification.) For VarChar, a precision of 0 means that the user requested {{VARCHAR}}, while a precision of, say, 10 means the user requested {{VARCHAR(10}}. The scale is never valid for {{VARCHAR}}, it is an artifact of the incorrect way the above code was written. The Protobuf issue is that, unlike a regular Java object, if we never actually write to the precision field, then the value is unset. If we write, even if we write 0, the value is set. We certainly don't want to litter our code with things like: {code:java} if (precision != 0) { schemaBuilder.setPrecision(precision); } {code} So, we should ask if the precision is set and non-zero. In fact, the type formatting code should not even be in the REST API. The proper place for it is in {{Types}}. In fact, that class already has the desired function: {{getExtendedSqlTypeName()}}. However, this function only formats decimals; we need to add a case clause for VARCHAR. That said, I actually have not seen any places in Drill where we set or use the VARCHAR width. So, no point in trying to format it. In this case, you can just use {{getExtendedSqlTypeName()}} directly as-is. Please file a separate JIRA for the UDF issue. Please provide an attachment or link to a sample UDF. I'll see if I can track down that CSV-specific issue in case it relates to the EVF. was (Author: paul.rogers): [~cgivre], the problem here is that the code shown earlier is counting on a Protobuf implementation detail that is not actually a part of the Drill schema specification (to the degree there is such a specification.) For VarChar, a precision of 0 means that the user requested {{VARCHAR}}, while a precision of, say, 10 means the user requested {{VARCHAR(10}}. The scale is never valid for {{VARCHAR}}, it is an artifact of the incorrect way the above code was written. The Protobuf issue is that, unlike a regular Java object, if we never actually write to the precision field, then the value is unset. If we write, even if we write 0, the value is set. We certainly don't want to litter our code with things like: ``` if (precision != 0) { schemaBuilder.setPrecision(precision); } ``` So, we should ask if the precision is set and non-zero. In fact, the type formatting code should not even be in the REST API. The proper place for it is in {{Types}}. In fact, that class already has the desired function: {{getExtendedSqlTypeName()}}. However, this function only formats decimals; we need to add a case clause for VARCHAR. That said, I actually have not seen any places in Drill where we set or use the VARCHAR width. So, no point in trying to format it. In this case, you can just use {{getExtendedSqlTypeName()}} directly as-is. Please file a separate JIRA for the UDF issue. Please provide an attachment or link to a sample UDF. I'll see if I can track down that CSV-specific issue in case it relates to the EVF. > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following metadata: > {code:sql} > SELECT * FROM dfs.test.`domains.csvh` LIMIT 1 > {code} > {code:json} > { > "queryId": "22eee85f-c02c-5878-9735-091d18788061", > "columns": [ > "domain" > ], > "rows": [} > { "domain": "thedataist.com" } ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > "queryState": "COMPLETED", > "attemptedAutoLimit": 0 > } > {code} > There are two issues here: > 1. VARCHAR now has precision > 2. There are twice as many columns as there should be. > Additionally, if you query a regular CSV, without the columns extracted, you > get the following: > {code:json} > "rows": [ > { > "columns": "[\"ACCT_NUM\",\"PRODUCT\",\"MONTH\",\"REVENUE\"]" } > ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > {code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16871990#comment-16871990 ] Paul Rogers edited comment on DRILL-7308 at 6/25/19 5:27 AM: - The width issue appears to have been introduced with this commit: "DRILL-6847: Add Query Metadata to RESTful Interface" (which, ahem, [~cgivre], was your PR...). In {{WebUserConnection}}: {code:java} //For DECIMAL type if (col.getType().hasPrecision()) { dataType.append("("); dataType.append(col.getType().getPrecision()); if (col.getType().hasScale()) { dataType.append(", "); dataType.append(col.getType().getScale()); } dataType.append(")"); } else if (col.getType().hasWidth()) { //Case for VARCHAR columns with specified width dataType.append("("); dataType.append(col.getType().getWidth()); dataType.append(")"); } {code} I did not debug the code, but it appears that {{hasPrecision()}} and {{hasScale()}} simply report if the field is set; it does *not* tell us if the field is zero. Also, about a year or so ago, Drill moved {{VARCHAR}} width to the precision field, so the supposed {{VARCHAR}} code block is a no-op. The correct code would be something like: {code:java} //For DECIMAL and VARCHAR types if (col.getType().hasPrecision() && col.getType().getPrecision() > 0) { dataType.append("("); dataType.append(col.getType().getPrecision()); if (col.getType().hasScale() && col.getType().getScale() > 0) { {code} was (Author: paul.rogers): The width issue appears to have been introduced with this commit: "DRILL-6847: Add Query Metadata to RESTful Interface" (which, ahem, [~cgivre], was your PR...): {code:java} //For DECIMAL type if (col.getType().hasPrecision()) { dataType.append("("); dataType.append(col.getType().getPrecision()); if (col.getType().hasScale()) { dataType.append(", "); dataType.append(col.getType().getScale()); } dataType.append(")"); } else if (col.getType().hasWidth()) { //Case for VARCHAR columns with specified width dataType.append("("); dataType.append(col.getType().getWidth()); dataType.append(")"); } {code} I did not debug the code, but it appears that {{hasPrecision()}} and {{hasScale()}} simply report if the field is set; it does *not* tell us if the field is zero. Also, about a year or so ago, Drill moved {{VARCHAR}} width to the precision field, so the supposed {{VARCHAR}} code block is a no-op. The correct code would be something like: {code:java} //For DECIMAL and VARCHAR types if (col.getType().hasPrecision() && col.getType().getPrecision() > 0) { dataType.append("("); dataType.append(col.getType().getPrecision()); if (col.getType().hasScale() && col.getType().getScale() > 0) { {code} > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following metadata: > {code:sql} > SELECT * FROM dfs.test.`domains.csvh` LIMIT 1 > {code} > {code:json} > { > "queryId": "22eee85f-c02c-5878-9735-091d18788061", > "columns": [ > "domain" > ], > "rows": [} > { "domain": "thedataist.com" } ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > "queryState": "COMPLETED", > "attemptedAutoLimit": 0 > } > {code} > There are two issues here: > 1. VARCHAR now has precision > 2. There are twice as many columns as there should be. > Additionally, if you query a regular CSV, without the columns extracted, you > get the following: > {code:json} > "rows": [ > { > "columns": "[\"ACCT_NUM\",\"PRODUCT\",\"MONTH\",\"REVENUE\"]" } > ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > {code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16871992#comment-16871992 ] Paul Rogers edited comment on DRILL-7308 at 6/25/19 4:41 AM: - This issue points out a unfortunate reality (IMHO): lack of unit tests for the REST API. We have nothing, other than vigilent users, to track down issues such as this one. I believe that unit tests can be easily created: use a {{ClusterTest}} and set the config\(?) option to enable the web server. Use a web client of some sort to fire a request. Either compare the results against a golden file, or just test for the bits of interest (such as, for DRILL-6847, test against each type and mode, and with and without width/precision, and verify just that part of the result. The unit test would also allow very easy debugging. It seems the best we can do at present is build all of Drill, start it, and connect a remote debugger. This is so cumbersome that folks will avoid stepping through code to see if it works. was (Author: paul.rogers): This issue points out a unfortunate reality (IMHO): lack of unit tests for the REST API. We have nothing, other than vigilent users, to track down issues such as this one. I believe that unit tests can be easily created: use a {{ClusterTest}} and set the config(?) option to enable the web server. Use a web client of some sort to fire a request. Either compare the results against a golden file, or just test for the bits of interest (such as, for DRILL-6847, test against each type and mode, and with and without width/precision, and verify just that part of the result. The unit test would also allow very easy debugging. It seems the best we can do at present is build all of Drill, start it, and connect a remote debugger. This is so cumbersome that folks will avoid stepping through code to see if it works. > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following metadata: > {code:sql} > SELECT * FROM dfs.test.`domains.csvh` LIMIT 1 > {code} > {code:json} > { > "queryId": "22eee85f-c02c-5878-9735-091d18788061", > "columns": [ > "domain" > ], > "rows": [} > { "domain": "thedataist.com" } ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > "queryState": "COMPLETED", > "attemptedAutoLimit": 0 > } > {code} > There are two issues here: > 1. VARCHAR now has precision > 2. There are twice as many columns as there should be. > Additionally, if you query a regular CSV, without the columns extracted, you > get the following: > {code:json} > "rows": [ > { > "columns": "[\"ACCT_NUM\",\"PRODUCT\",\"MONTH\",\"REVENUE\"]" } > ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > {code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (DRILL-7308) Incorrect Metadata from text file queries
[ https://issues.apache.org/jira/browse/DRILL-7308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16871945#comment-16871945 ] Paul Rogers edited comment on DRILL-7308 at 6/25/19 3:08 AM: - According to the screen shot, this is the REST API, method POST with query as payload, URL is {{http::/query.json}}. was (Author: paul.rogers): I presume this is the REST API? Please specify the URL used to do the query. > Incorrect Metadata from text file queries > - > > Key: DRILL-7308 > URL: https://issues.apache.org/jira/browse/DRILL-7308 > Project: Apache Drill > Issue Type: Bug > Components: Metadata >Affects Versions: 1.17.0 >Reporter: Charles Givre >Priority: Major > Attachments: Screen Shot 2019-06-24 at 3.16.40 PM.png, domains.csvh > > > I'm noticing some strange behavior with the newest version of Drill. If you > query a CSV file, you get the following metadata: > {code:sql} > SELECT * FROM dfs.test.`domains.csvh` LIMIT 1 > {code} > {code:json} > { > "queryId": "22eee85f-c02c-5878-9735-091d18788061", > "columns": [ > "domain" > ], > "rows": [} > { "domain": "thedataist.com" } ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > "queryState": "COMPLETED", > "attemptedAutoLimit": 0 > } > {code} > There are two issues here: > 1. VARCHAR now has precision > 2. There are twice as many columns as there should be. > Additionally, if you query a regular CSV, without the columns extracted, you > get the following: > {code:json} > "rows": [ > { > "columns": "[\"ACCT_NUM\",\"PRODUCT\",\"MONTH\",\"REVENUE\"]" } > ], > "metadata": [ > "VARCHAR(0, 0)", > "VARCHAR(0, 0)" > ], > {code} -- This message was sent by Atlassian JIRA (v7.6.3#76005)