以导出多表数据为例看 ABAP 如何逐步优化
问：老师，我要导出SAP的好多个表的全部表内容，请问有没有合适的程序给我用用呢？
答：这个程序很简单，随便写一下就行了：
REPORT zexport_table NO STANDARD PAGE HEADING.TABLES dd02l.DATA: gt_tabs TYPE TABLE OF tabname WITH HEADER LINE.DATA: go_table TYPE REF TO data.DATA: gv_lfile TYPE string.FIELD-SYMBOLS: <gt_table> TYPE STANDARD TABLE.
SELECT-OPTIONS s_tabnam FOR dd02l-tabname.START-OF-SELECTION. SELECT tabname INTO TABLE gt_tabs FROM dd02l WHERE tabname IN s_tabnam AND as4local EQ 'A' AND tabclass IN ('TRANSP','CLUSTER','POOL','VIEW'). LOOP AT gt_tabs. CREATE DATA go_table TYPE TABLE OF (gt_tabs). ASSIGN go_table->* TO <gt_table>.
SELECT * FROM (gt_tabs) INTO TABLE <gt_table>. CHECK sy-subrc = 0.gv_lfile = 'D:\tmp\' && gt_tabs && '.txt'. CALL FUNCTION 'GUI_DOWNLOAD' EXPORTING filename = gv_lfile filetype = 'DAT' confirm_overwrite = ' ' TABLES data_tab = <gt_table> EXCEPTIONS OTHERS = 22. IF sy-subrc NE 0. MESSAGE i000(oo) WITH '文件下载错误：' gt_tabs. ENDIF. ENDLOOP.
问：太好了！小小的代码大大的能量， 果然导出来了。 可是导出的文件没有列名呢，对数非常不方便啊！
答：那就加上字段名作为列名吧，除了程序中用的的方法，也可以在表DD03L中取列名。
TABLES dd02l.DATA: gt_tabs TYPE TABLE OF tabname WITH HEADER LINE.DATA: go_table TYPE REF TO data.DATA: gv_lfile TYPE string.DATA: go_tdescr TYPE REF TO cl_abap_tabledescr, go_sdescr TYPE REF TO cl_abap_structdescr, gs_comps LIKE LINE OF cl_abap_structdescr=>components, gt_field TYPE TABLE OF dd03l.
FIELD-SYMBOLS: <gt_table> TYPE STANDARD TABLE.
SELECT-OPTIONS s_tabnam FOR dd02l-tabname.START-OF-SELECTION. SELECT tabname INTO TABLE gt_tabs FROM dd02l WHERE tabname IN s_tabnam AND as4local EQ 'A' AND tabclass IN ('TRANSP','CLUSTER','POOL','VIEW'). LOOP AT gt_tabs. CREATE DATA go_table TYPE TABLE OF (gt_tabs). ASSIGN go_table->* TO <gt_table>.
SELECT * FROM (gt_tabs) INTO TABLE <gt_table>. CHECK sy-subrc EQ 0.CLEAR gt_field[]. go_tdescr ?= cl_abap_typedescr=>describe_by_data( <gt_table> ). go_sdescr ?= go_tdescr->get_table_line_type( ). LOOP AT go_sdescr->components INTO gs_comps. APPEND gs_comps-name TO gt_field. ENDLOOP.
gv_lfile = 'D:\tmp\' && gt_tabs && '.txt'. CALL FUNCTION 'GUI_DOWNLOAD' EXPORTING filename = gv_lfile filetype = 'DAT' confirm_overwrite = ' ' TABLES data_tab = <gt_table> fieldnames = gt_field EXCEPTIONS OTHERS = 22. IF sy-subrc NE 0. MESSAGE i000(oo) WITH '文件下载错误：' gt_tabs. ENDIF. ENDLOOP.
问：太好啦！可是这种前台导出太慢了，导出一个几十万数据的表都要好久呢，等的花儿都谢了！
答：确实是这样，谁知道你要导出这么大的表呢！前台用GUI_DOWNLOAD导出数据性能是比较差的，我给你改成后台运行吧，然后把数据存到服务器上，再从服务器下载就非常非常快了。程序执行完在JOB日志可以看的文件名，然后可以用CG3Y的BIN格式下载，或者写一个程序下载。文件放到应用服务器的DIR_HOME目录哦，用AL11可以看到存放的文件。
TABLES dd02l.DATA: gt_tabs TYPE TABLE OF tabname WITH HEADER LINE.DATA: go_table TYPE REF TO data.DATA: gv_line TYPE string.DATA: gv_afile TYPE string.DATA: go_tdesc TYPE REF TO cl_abap_tabledescr, go_sdesc TYPE REF TO cl_abap_structdescr, gs_comps LIKE LINE OF cl_abap_structdescr=>components.
FIELD-SYMBOLS: <gt_table> TYPE STANDARD TABLE,<gs_table>,<gv_fld>.
SELECT-OPTIONS s_tabnam FOR dd02l-tabname.START-OF-SELECTION. SELECT tabname INTO TABLE gt_tabs FROM dd02l WHERE tabname IN s_tabnam AND as4local EQ 'A' AND tabclass IN ('TRANSP','CLUSTER','POOL','VIEW'). LOOP AT gt_tabs. CREATE DATA go_table TYPE TABLE OF (gt_tabs). ASSIGN go_table->* TO <gt_table>.
SELECT * FROM (gt_tabs) INTO TABLE <gt_table>. CHECK sy-subrc EQ 0.CLEAR gv_line. go_tdesc ?= cl_abap_typedescr=>describe_by_data( <gt_table> ). go_sdesc ?= go_tdesc->get_table_line_type( ). LOOP AT go_sdesc->components INTO gs_comps. gv_line = gv_line && %_horizontal_tab && gs_comps-name. ENDLOOP. SHIFT gv_line.
gv_afile = gt_tabs && '_' && '_' && sy-dbcnt && '.TXT'. OPEN DATASET gv_afile FOR OUTPUT IN TEXT MODE ENCODING UTF-8 WITH WINDOWS LINEFEED. TRANSFER gv_line TO gv_afile.
LOOP AT <gt_table> ASSIGNING <gs_table>. CLEAR gv_line. DO. ASSIGN COMPONENT sy-index OF STRUCTURE <gs_table> TO <gv_fld>. IF sy-subrc NE 0. EXIT. ENDIF.gv_line = gv_line && %_horizontal_tab. gv_line = gv_line && <gv_fld>. ENDDO. SHIFT gv_line. TRANSFER gv_line TO gv_afile. ENDLOOP. CLOSE DATASET gv_afile. MESSAGE s000(oo) WITH gv_afile. ENDLOOP.
问：太棒了老师！运行快多了。可是我在下载一个超大表的时候程序崩掉了呢，数据量太大了，把内表撑爆了！
答：你都是导出啥表啊竟然能撑爆内表！到目前为止呢，我们都是一次性取出数据库表的数据到内表，如果数据量非常大，超过了系统参数所限定的内存大小就DUMP了。这样吧，我们取数据改为使用游标的方式，每次取出来100000条数据，等保存完后再继续取下一个100000条，如此往复，就不怕内表撑爆了。另外，我知道你还会遇到另外的问题，所以咱们先把文件在程序内转成XSTRING再存放，至于为啥等你遇到下一个问题就知道啦
TABLES dd02l.DATA gt_tabs TYPE TABLE OF tabname WITH HEADER LINE.DATA go_table TYPE REF TO data.DATA gv_cursor TYPE cursor.DATA gv_afile TYPE string.DATA gv_header TYPE string.DATA gv_line TYPE string.DATA gv_lines TYPE string.DATA gv_xstr TYPE xstring.DATA go_tdesc TYPE REF TO cl_abap_tabledescr.DATA go_sdesc TYPE REF TO cl_abap_structdescr.DATA gs_comps LIKE LINE OF cl_abap_structdescr=>components.
FIELD-SYMBOLS: <gt_table> TYPE STANDARD TABLE,<gs_table>,<gv_fld>.
SELECT-OPTIONS s_tabnam FOR dd02l-tabname.START-OF-SELECTION. SELECT tabname INTO TABLE gt_tabs FROM dd02l WHERE tabname IN s_tabnam AND as4local EQ 'A' AND tabclass IN ('TRANSP','CLUSTER','POOL','VIEW'). LOOP AT gt_tabs. CREATE DATA go_table TYPE TABLE OF (gt_tabs). ASSIGN go_table->* TO <gt_table>.
CLEAR gv_header. go_tdesc ?= cl_abap_typedescr=>describe_by_data( <gt_table> ). go_sdesc ?= go_tdesc->get_table_line_type( ). LOOP AT go_sdesc->components INTO gs_comps. gv_header = gv_header && %_horizontal_tab && gs_comps-name. ENDLOOP. SHIFT gv_header.
OPEN CURSOR gv_cursor FOR SELECT * FROM (gt_tabs). DO. FETCH NEXT CURSOR gv_cursor INTO TABLE <gt_table> PACKAGE SIZE 100000. IF sy-subrc NE 0. EXIT. ENDIF.
gv_afile = gt_tabs && '_' && sy-index && '_' && sy-dbcnt && '.TXT'.
OPEN DATASET gv_afile FOR OUTPUT IN BINARY MODE. gv_lines = gv_header && %_cr_lf. LOOP AT <gt_table> ASSIGNING <gs_table>. CLEAR gv_line. DO. ASSIGN COMPONENT sy-index OF STRUCTURE <gs_table> TO <gv_fld>. IF sy-subrc NE 0. EXIT. ENDIF.
gv_line = gv_line && %_horizontal_tab. gv_line = gv_line && <gv_fld>. ENDDO. SHIFT gv_line. gv_line = gv_line && %_cr_lf. gv_lines = gv_lines && gv_line. ENDLOOP.
CALL FUNCTION 'SCMS_STRING_TO_XSTRING' EXPORTING text = gv_lines mimetype = 'UTF-8' IMPORTING buffer = gv_xstr.TRANSFER gv_xstr TO gv_afile. CLOSE DATASET gv_afile. MESSAGE s000(oo) WITH gv_afile. ENDDO. CLOSE CURSOR gv_cursor. ENDLOOP.
问：哇哈哈，果然内表不会爆掉了！不过我真的又遇到问题了呢，文件太大了，把应用服务器给撑爆了，磁盘空间都没有啦！老师你刚才说会遇到新的问题，是不是这个问题呢？
答：就是这个问题，哈哈哈，我们之所以把数据转为XSTRING再存放就是为了解决这个问题。我们知道SAP数据这种二维表的数据冗余是非常严重的，稍微压缩一下就能有一个很好的压缩比，有的时候能压缩到原大小的十分之一甚至百分之一呢！
TABLES dd02l.DATA gt_tabs TYPE TABLE OF tabname WITH HEADER LINE.DATA go_zip TYPE REF TO cl_abap_zip.DATA go_table TYPE REF TO data.DATA gv_cursor TYPE cursor.DATA gv_afile TYPE string.DATA gv_fname TYPE string.DATA gv_head TYPE string.DATA gv_line TYPE string.DATA gv_lines TYPE string.DATA gv_xstr TYPE xstring.DATA go_tdesc TYPE REF TO cl_abap_tabledescr.DATA go_sdesc TYPE REF TO cl_abap_structdescr.DATA gs_comps LIKE LINE OF cl_abap_structdescr=>components.
FIELD-SYMBOLS: <gt_table> TYPE STANDARD TABLE,<gs_table>,<gv_fld>.
SELECT-OPTIONS s_tabnam FOR dd02l-tabname.START-OF-SELECTION. SELECT tabname INTO TABLE gt_tabs FROM dd02l WHERE tabname IN s_tabnam AND as4local EQ 'A' AND tabclass IN ('TRANSP','CLUSTER','POOL','VIEW'). LOOP AT gt_tabs. CREATE DATA go_table TYPE TABLE OF (gt_tabs). ASSIGN go_table->* TO <gt_table>.
CLEAR gv_head. go_tdesc ?= cl_abap_typedescr=>describe_by_data( <gt_table> ). go_sdesc ?= go_tdesc->get_table_line_type( ). LOOP AT go_sdesc->
...
